Anthropic Release Notes

Added

• Added a diff panel that opens beside the conversation in fullscreen mode and shows your uncommitted changes as Claude edits; toggle it with /diff
• Added a likely cause for prompt-cache misses (e.g. tool definitions or system prompt changed, idle past the TTL) to /cost and the status line's prompt_cache field
• Added /reload-plugins to headless sessions, so it appears in the Claude Code Desktop and SDK command lists
• Added a text form of /advisor (/advisor, /advisor <model>, /advisor off) for the desktop app, Remote Control, and other headless (-p/Agent SDK) sessions
• Added oidc.scope_on_refresh to the Claude apps gateway for IdPs that return an id_token on refresh only when asked for openid again
• Added Claude apps gateway support for newer Claude Desktop keys in desktop policy blocks, including userPluginMarketplacesEnabled and userPluginUploadsEnabled

Fixed

• Fixed Edit/Write/Read permission rules whose path contains parentheses being dropped as invalid or ignored by the Bash sandbox, which left "read-only" folders writable
• Fixed one file permission rule with an uncompilable pattern (e.g. an unclosed [) making every file edit fail with Invalid regular expression; such a deny rule now guards the literal path it spells
• Fixed Bash permission checks auto-approving zsh commands that hide a command substitution in a REPORTTIME, REPORTMEMORY or DIRSTACKSIZE assignment; these now prompt for approval
• Fixed Bedrock model discovery, token counting and AWS SSO/STS credential calls failing with "unable to get local issuer certificate" when the corporate root CA is only in the OS certificate store
• Fixed permissions.blockReadsOutsideWorkingDirectories on macOS hiding the user's git config from sandboxed git and hiding a worktree-isolated sub-agent's own checkout
• Fixed managed settings not loading for claude.ai Enterprise/Team users who also had a leftover API key from an earlier /login
• Fixed /status listing a signed-in claude.ai account and a configured API key as if both were in effect; the credential not in use is now marked
• Fixed managed skillOverrides entries keyed on a bundled skill's alias (e.g. checkup for /doctor) not applying, and Skill(name) deny rules not covering a nested skill listed as <dir>:name
• Fixed model: fable agents ignoring the [1m] tag on an ANTHROPIC_DEFAULT_FABLE_MODEL pin and silently running with a 200K context window
• Fixed the /model picker not showing Fable 5.1 for organizations that can use it, which was only accepted when typed as /model claude-fable-5-1
• Fixed prompt caching on Claude Fable 5.1 not covering the context attached after tool results, so it was re-sent as uncached input on every tool-call turn
• Fixed model switching staying blocked for the rest of the session after a plugin hook load failure; each switch now re-checks and the refusal names the cause
• Fixed model switching being blocked for the session when an organization-managed plugin's marketplace could not be loaded
• Fixed SDK-provided MCP servers (e.g. Desktop connectors) sometimes missing from the first turn and only appearing on the next one
• Fixed Claude in Chrome tools failing with "Not connected" mid-task in cloud-hosted claude.ai sessions when a connector was added or removed
• Fixed flags, joined emoji and accented letters splitting across wrapped lines, and stale text staying on screen when a flag or joined emoji falls in the terminal's last two columns (now shown as …)
• Fixed Remote Control accepting a model pick that is not a valid model name; it is now refused with an error instead of failing on the next message
• Fixed /rewind and --rewind-files reporting success when checkpoint backup files were missing and nothing was actually restored
• Fixed /rewind leaving stale file-read tracking from the rewound-away turns, which caused "File unchanged since last read" stubs and full-file re-injection after external edits
• Fixed -p --resume/--continue (as used by the desktop app) failing on every retry once a session's worktree directory lost its git metadata; it now fails once, then resumes without the worktree
• Fixed a subagent that resumed another agent via SendMessage never being woken by that agent's completion (the notification went to the main conversation instead)
• Fixed agent teams: an in-process teammate's transcript losing messages, or going blank, during long API retry waits (e.g. under CLAUDE_CODE_RETRY_WATCHDOG) as retry notices evicted real messages
• Fixed a session that moved to the background appearing twice in ListAgents (once as a phantom "interactive" twin with the same name) and receiving SendMessage deliveries in the viewer
• Fixed intermittent "task output swap refused" errors when many sessions share a project directory
• Fixed Ctrl+Z in fullscreen leaving the shell on the alternate screen, drawn over the paused interface
• Fixed Workflow tool subagents being restarted as stalled while a long context compaction was still in progress
• Fixed plugins from a URL marketplace failing to install with "marketplace entry path does not stay inside the marketplace directory" when a host app (e.g. Claude Desktop) stores it as a directory
• Fixed an extra browser tab opening when an artifact is published in a session you're driving from claude.ai, the desktop app, or mobile (Remote Control)
• Fixed the Artifact tool's first call failing with an "Invalid tool parameters" validation error in some Cowork sessions
• Fixed IDE line selections being dropped when running a skill or slash command (the "N lines selected" context now reaches Claude)
• Fixed repository detection for GitLab projects in nested subgroups (e.g. gitlab.com/group/subgroup/project)
• Fixed owner/repo#123 issue references in rendered output linking to github.com when working in a GitLab repository; they now link to the gitlab.com issue
• Fixed Glob/Grep: Fixed the search path being probed on disk before the permission check; a missing path is now reported after permission is decided, as Read does
• Reverted the 2.1.259 change applying Read() deny rules to Bash arguments; it denied npm run build under a Read(./**/build/**) rule in every mode and made cd … && grep prompt even in auto mode

Improved

• Improved structured output: Workflow agent({schema}) rejects a JSON Schema that can never be satisfied up front, and retry-cap errors now include the last validation failure
• Improved deleting a background session whose worktree has unpushed commits: the message now names the branch and commit count, and deleting again discards the worktree
• Improved the Claude apps gateway's refresh-failure log to name the step that failed
• Improved idle CPU usage of non-interactive (-p / SDK) sessions
• Improved the Claude apps gateway on Amazon Bedrock: input tokens for an aborted request are now counted with AWS's free CountTokens API (grant bedrock:CountTokens) instead of a one-token request
• Improved the settings error for rules such as Edit(C:\dir\(name)\**), where \( is read as an escaped parenthesis rather than a path separator, to suggest an unambiguous spelling
• Improved auto-compact for 1M-context models: Opus and Fable sessions now compact shortly before the 1M-token limit, and recovery compaction on very large contexts no longer times out at 10 minutes
• Improved /ultrareview and claude ultrareview to wait up to 45 minutes (previously 30) for long-running cloud reviews
• Improved /effort on Claude Fable 5.1 so changing effort mid-session no longer invalidates the prompt cache

Updated

• Updated the bundled claude-api skill so its Go, Java, and C# samples use current-generation model IDs, and clarified that cheaper worker or sub-agent models should be current-generation too

Changed

• Changed ctrl+l / cmd+k in fullscreen mode to clear the transcript view like a terminal clear; scroll up to see earlier messages
• Changed permission rules with text after the closing parenthesis (e.g. Bash(ls) x), which never matched anything, to be reported as invalid settings instead of being silently ignored
• Changed server-managed settings so a managed CLAUDE.md (claudeMd) no longer triggers the security approval dialog; hooks, shell-command, sandbox, and unsafe env settings still require approval
• Changed Claude in Chrome to follow your organization's Claude in Chrome admin setting; when an admin turns it off, --chrome, /chrome and the browser tools are unavailable
• Changed Claude apps gateway to send orgPluginSettings in the list form read by Claude Desktop 1.15200.0 and later; older desktops ignore it
• Changed Claude apps gateway to also refuse to start, naming the field, when a desktop policy misspells a field in a nested object of a managedMcpServers or orgPluginSettings entry
• Changed commands typed at the ! bash-mode prompt to run outside the sandbox even when strict sandbox mode (sandbox.allowUnsandboxedCommands: false) is on, like typing into your own terminal
• Changed self-hosted runner --kill-session-after-min to release a session that is only waiting on its user (paused, resumable on the next message) instead of killing it and reporting a failure

Removed

• Removed the one-hour time limit on background commands started by subagents; they now run until they exit or are stopped, matching the main session

VSCode

• [VSCode] Added the selected effort level to the footer model pill, fixed a stale effort level after switching models, and returned the footer pills to their earlier compact size
• [VSCode] Added Open and Closed to the session list's status filter menu
• [VSCode] Fixed the welcome screen disappearing in a new session when Remote Control turns on automatically
• [VSCode] Fixed the session history picker loading a session a second time when it is already open in another tab; it now switches to that tab
• [VSCode] Fixed the session tab's Rename command silently doing nothing while the tab's view was reloading; it now always applies
• [VSCode] Fixed a half-finished message, an empty tool card or an extra "Thought for" line staying on screen after Claude Code retried a dropped response
• [VSCode] Fixed "Enable Remote Control for all sessions" not applying to a session tab that was still starting when the toggle was flipped

Added

• Added managedMcpServers managed setting: organizations can provide HTTP/SSE MCP servers to every user (same entry shape as .mcp.json); entries that name a command to run are skipped
• Added --permission-prompts none for unattended headless hosts: anything that would prompt is denied automatically while the active permission mode (including auto mode) keeps deciding
• Added recognition of glab mr create/merge/close/reopen/note/update so GitLab merge requests show as MR !N in the collapsed tool summary and refresh the footer MR badge
• Added --json to claude plugin validate for a machine-readable validation report

Fixed

• Fixed concurrent sessions silently reverting each other's ~/.claude.json changes — workspace trust no longer resets and MCP/project state is no longer lost when running many sessions at once
• Fixed a conversation whose thinking was rejected once being rejected again on every later turn
• Fixed Bash Read() deny rules not covering files given as option values (--ignore-revs-file=.env, -f.env, @file), git diff/git grep file operands, or cd DIR && cat FILE compounds; grep -r/cp -r over a directory holding a denied file now asks
• Fixed the prompt cache being invalidated when the OAuth token refreshed in sessions with telemetry disabled
• Fixed fullscreen mode showing a blank conversation after a long turn with hundreds of tool calls
• Fixed auto mode running a turn on a model it doesn't support when a command or skill's frontmatter model: named one; the turn now keeps the session model
• Fixed CLAUDE_CODE_MAX_CONTEXT_TOKENS being ignored for Vertex-style model IDs (@YYYYMMDD suffix) of model versions Claude Code doesn't recognize
• Fixed the live output preview of a running shell command hiding its newest lines when an earlier line wrapped
• Fixed a background GitHub connection check that ran on every launch for claude.ai users; the result is now remembered across launches
• Fixed --resume failing (and --continue opening an empty conversation) when a saved session contains an attachment entry with no payload
• Fixed frontmatter model: on custom commands and skills being ignored in interactive sessions
• Fixed Artifact publishing failing once with an "unexpected parameter note" error in conversations continued from an older version
• Fixed managed forceRemoteSettingsRefresh being ignored at startup when a policy helper configured by MDM or the managed settings file had already run
• Fixed worktree isolation refusing hook-created worktrees on machines where git rev-parse fails with a message other than "not a git repository"
• Fixed OpenTelemetry metrics and events from cloud sessions missing the user.email, organization.id, and user.account_uuid attributes
• Fixed MCP servers that disconnect while their tools are being listed at startup showing as connected with no tools instead of reporting the error
• Fixed the file edit permission dialog sometimes showing a changed line cut short with no indication
• Fixed repository detection dropping a known repo identity after a transient git probe failure
• Fixed managed settings silently going unenforced when the managed-settings file, a drop-in, the MDM plist, or the HKLM value cannot be parsed: Claude Code now refuses to start and names the source
• Fixed Stop not actually stopping background agents and workflows in remote-control sessions: killed tasks now stay visible and re-stoppable until their processes exit
• Fixed resuming a workflow run while its previous stopped run was still exiting, which could run duplicate copies of its agents
• Fixed marketplace repo URLs on github.com with a trailing slash or dangling ?/# producing an unusable .git clone URL
• Fixed blocking Stop hooks causing the turn after a block to lose the model's reasoning from that turn and, on some models, miss the prompt cache
• Fixed remote (claude.ai) sessions taking 60 seconds to start a turn after a browser-hosted MCP server's page had gone away
• Fixed worktree-isolated sessions refusing common Bash loops, xargs pipelines and launcher-wrapped commands that cannot reach the main checkout

Improved

• Improved terminal resize and first-render performance for long responses by reusing text measurements
• Improved /workflows agent detail: JSON outcomes are pretty-printed with syntax colors and real line breaks, and long outcomes fold behind an expand toggle
• Improved headless/SDK session start: the first turn begins up to 50 ms sooner when MCP servers finish connecting
• Improved /install-github-app to explain it is GitHub-only and point to the GitLab CI/CD docs when run inside a GitLab repository
• Improved nested background subagent results to be saved in the parent subagent's transcript, so resumed subagents keep them and shared transcripts show the delivery

Changed

• Changed allowedMcpServers to govern only servers users add: a literal managed-mcp.json server your allowlist used to filter out now loads on upgrade; use deniedMcpServers to keep it off

VSCode

• [VSCode] Added an Active quick filter and a status filter menu (Needs input, Working, Completed) to the session list sidebar

Fixed

• Fixed remote and scheduled sessions doing nothing after a connector-tool permission prompt was approved while the session was paused

What's in the blueprint

Many of the world’s largest retailers, marketplaces, e-commerce platforms, and travel companies use Claude to build agents that make shopping easier. Enterprise customers like Shopify, Priceline, and others have agents that let consumers use AI to search for what they want in plain language, find it, compare it, and buy it.

Today, we're launching a blueprint to help build commerce agents on Claude. It contains the harnesses, patterns, and guardrails an engineering team needs to get a commerce agent running in days, with reference implementations of a shopping agent and a merchant agent for retail, travel, telecom, and ticketing platforms. It also includes a Claude Code plugin to get you started.

The code deploys where you already build with Claude, including the Claude API, Amazon Bedrock, Microsoft Foundry, or Google Cloud Vertex AI. You can also work with our solutions and ecosystem partners such as Accenture, Mastercard, and Visa, who are working with us to enable clients and merchant communities to leverage the blueprints.

It’s available today, with live demos for each vertical and an engineering deep-dive on how it was built, just in time for holiday season planning.

The repository contains complete, working implementations of a shopping agent and merchant agent that can be built using the Messages API, Agent SDK, or Claude Managed Agents (beta). You can see them running in a self-guided demo before writing any code, and then work with Claude Code to customize them to your catalogs, policies, brand, and more.

The shopping agent

The shopping agent lives inside your app or website. The blueprint includes the integration points for catalog, cart, checkout, customer preferences, and order history, and leaves payment to you, whether that is your existing checkout or an agentic payments provider.

A customer can say “I need a tent, sleeping bag, and stove for a weekend trip with two kids,” and the agent can take it from there. Here’s what it can do:

• Search the catalog and assemble the right set of items, including multi-item requests.
• Remember the customer's preferences and tailor what it suggests.
• Show products, comparisons, and the cart right in the conversation, not just as text.
• Build the cart and hand it to checkout.
• Answer customer service questions in the same conversation, like where an order is, how to return or exchange an item, and what the refund policy says, instead of sending the customer to a support page.

The agent features guardrails designed to constrain prices and products to actual catalog data, and avoids manipulative upsell patterns. In the repository, these are skills and tools for catalog search, multi-item planning, deep research, personalization, customer care, and in-conversation UI.

The merchant agent

The merchant agent supports the people running the store. A user can ask “what should we discount to clear last season’s inventory?” and get an answer based on their own data. Here’s what it can do:

• Answer questions about sales performance like what's selling and what isn't.
• Track inventory and proactively flag problems, like an item about to sell out before a promotion starts.
• Recommend pricing and promotions based on the store's own sales history.
• Draft marketing campaigns to move the products that need moving.

When the agent proactively suggests a change, a person approves it before anything goes live, meaning users get the final say while their agent watches the store. In the repository, these capabilities ship as skills for sales analytics, catalog and inventory management, marketing and promotions, and in-portal UI such as charts and dashboards.

Trusted across the industry

Companies that serve shoppers, travelers, subscribers, and merchants build and run agents on Claude. Here's what they have to say about building commerce agents with Claude:

“AI will fundamentally reshape commerce, but trust must remain at the center of every transaction. Merchants are telling us they want more control over how AI engages their customers. Our collaboration with Anthropic on their commerce blueprint helps bring together the intelligence of Claude with the trust, security, and global reach of the Visa network, empowering merchants to deliver better customer experiences while maintaining the relationships that drive their businesses forward.”
Jack Forestell, Chief Product and Strategy Officer

“Trust is the currency of commerce, and it is even more critical in the agentic era. With Anthropic's commerce blueprint, we're helping merchants build their own agents with Claude to drive their growth. By combining AI innovation with trusted payments and commerce infrastructure, we're helping connect consumers, merchants and AI agents securely, seamlessly and at scale.”
Sherri Haymond, Executive Vice President, Global Head of Digital Commercialization

“Commerce agents are quickly becoming a critical capability for organizations seeking to deliver the personalized, intelligent customer experiences that today’s consumers expect. Our latest research revealed that 85% are now open to collaboration with an AI agent and nearly three in four would trust a personal AI agent more than their best friend to make a purchase on their behalf. This is more than a shift in how people shop. Agentic commerce is rewriting the rules of brand value – fundamentally shaping what gets purchased, when, where and by whom. Anthropic’s commerce agent blueprint provides a proven starting point that can help organizations accelerate deployment and build differentiated experiences that increase customer satisfaction, loyalty, and growth. Combined with Accenture's deep retail and consumer goods expertise, we can help clients move from concept to production and realize value from agentic AI faster.”
Kath Gramling, Global Consumer Goods, Retail and Travel lead

“A trip is one of the most complex things a person buys: flights, hotels, cars, and dozens of options to weigh against each other. Penny, our AI assistant, navigates all of that in one conversation and surfaces the best options and best value. We built the latest generation of Penny on Claude because that kind of reasoning is exactly what Claude models are good at.”
Cobus Kok, Vice President, AI Experiences

“Millions of consumers, businesses and accountants run their finances on Intuit. Working with partners like Anthropic, we're building highly personalized experiences that provide customers with a clear understanding of what's shifting in their business and why, so they can take action with complete confidence. We are creating a financial system of intelligence by combining frontier AI reasoning, including Claude, with our proprietary data, capabilities, intelligence, and human expertise that powers the next level of prosperity for our customers.”
Chris Kasten, Intuit’s Chief Architect and SVP of Engineering, Platform and Development Xceleration Group

“We want our merchants to be everywhere customers are shopping, and increasingly that means a conversation with an agent. We're building on Anthropic's blueprints with a reference storefront implementation that connects them to a merchant's store through Catalog, UCP and Shop Sign-in. Merchants can use Claude to build agents that help customers find products, check out, and answer questions about their orders.”
Vanessa Lee, VP Product

“Commerce and stunning customer experiences require personalization, and every brand on Klaviyo sits on more customer data and decisions than any team could act on by hand. Claude closes that gap, turning consumer preferences and performance data into the insights, campaigns and personalization that drive revenue. That's why we keep building with Anthropic: agents do complex analysis, design and decision making, and businesses can focus on delighting customers.”
Andrew Bialecki, Founder and Co-CEO

“Wix’s mission has always been to make complex technology simple and accessible for our users. For merchants, that means providing powerful commerce capabilities without adding operational complexity, and agents are a natural next step. Our engineers had a working commerce agent taking prompts within fifteen minutes, and the pilot showed the potential of combining Anthropic’s AI capabilities with Wix’s commerce platform and deep expertise in commerce for SMB.”
Dror Zalika, Head of Commerce at Wix

“Our engineers had the blueprint running with no blockers; the setup worked exactly as documented. The practices it bakes in, from tool iteration limits to prompt caching, are the ones we recognized from building Zomato's own agent. Teams standing up their first agent on Claude will skip weeks of trial and error.”
Akhil Bansal, Senior Engineering Manager

“Our engineers had both commerce agents from Anthropic's blueprint running locally in well under an hour, with live conversations working on the first attempt. We ran the Claude Code workflow twice and got two different architectures back, each designed to what we'd asked for. For a team starting from scratch, that turns days of agent scaffolding into hours.”
Ashley Nader, Staff Product Manager

“Much of what we build at Square is about giving sellers time back, and agents are a big leap in our ability to do that. We're building agentic tools that watch sales, labor, and inventory and come back with real next steps, not just an answer, while keeping sellers in control. Trust is the hardest part of that work, and Claude helps us meet a high standard.”
Willem Avé, Head of Product

Getting started

The blueprint is available today. Contact our sales team to learn more, schedule a demonstration, or discuss how to implement for your organization.

1. Fork the repository at github.com/anthropics/commerce-agents.
2. Read the engineering deep-dive at claude.com/blog/the-anatomy-of-effective-commerce-agents.
3. See the vertical demos and request a working session at claude.com/solutions/commerce.

Register for our webinar to see the deep dive where we'll share live walkthroughs, demos, and cover how commerce builders can get the most out of Claude.

• We've launched Claude Fable 5.1 (claude-fable-5-1), the successor to Claude Fable 5 for long-running agentic coding, knowledge work, and research, alongside Claude Mythos 5.1 (claude-mythos-5-1) for Project Glasswing participants. Both models support a 1M token context window by default, 128k max output tokens, and always-on adaptive thinking, at $10 / $50 USD per MTok, the same as Claude Fable 5, with cache reads cut to $0.25 per MTok. Claude Fable 5.1 is available on the Claude API, Claude in Amazon Bedrock, Claude Platform on AWS, Claude on Google Cloud, and Claude in Microsoft Foundry. See What's new in Claude Fable 5.1 for capabilities, API changes, and migration guidance.

• Prompt cache reads on Claude Fable 5.1 and Claude Mythos 5.1 cost $0.25 USD per million tokens: 0.025x the base input price, compared with 0.1x on other models. Cache writes are unchanged. See Prompt caching pricing.

• On Claude Fable 5.1 and Claude Mythos 5.1, tool_choice types any and tool aren't supported and return a 400 error. auto and none are unchanged. To guarantee schema-conformant tool inputs, use strict tool use or structured outputs.

• Thinking blocks produced by Claude Fable 5.1 and Claude Mythos 5.1 are preserved only for the model that produced them or a newer one: earlier models can't read them, and the API drops one replayed to an earlier model. Claude Fable 5.1 accepts thinking blocks from Claude Opus 5, Claude Fable 5, Claude Mythos 5, and earlier Claude models. On Claude Fable 5.1, the API also checks that nothing before a block has changed: for new accounts created on or after August 31, 2026, replaying one after the system prompt, tools, or an earlier message changed returns a 400 error. With the thinking-binding-controls-2026-08-01 beta header, dropped blocks are reported in an input_transformations response field, and thinking.block_binding.prefix_mismatch_behavior chooses between rejecting and dropping blocks whose history changed. See Preserved thinking.

• Per-message effort changes are in beta on Claude Fable 5.1, Claude Mythos 5.1, and Claude Opus 5 on the Claude API. Add a role: "system" message with output_config.effort inside messages to change effort for later turns while preserving the prompt cache. Include the mid-conversation-output-config-2026-07-01 beta header in your requests. See Per-message effort.

• Turn-scoped system messages are in beta (mid-conversation-system-clear-at-2026-08-21 header). Set clear_at: "next_user_message" on a mid-conversation role: "system" message and it renders for the current turn only, then stays in the history at no token cost. Per-turn reminders don't accumulate and don't invalidate the prompt cache or later thinking blocks.

• thinking.display accepts a third value, "updates", in beta (thinking-display-updates-2026-08-18 header). Reasoning comes back with an empty thinking field, as under "omitted", and the short progress updates that Claude Fable 5.1, Claude Mythos 5.1, and Claude Fable 5 write between tool calls come back as text, at most one thinking block before a tool call. See Progress updates between tool calls.

• Text generated by Claude Fable 5.1 and Claude Mythos 5.1 carries Anthropic's text watermark, and supported image and video files that Claude produces through the code execution tool carry C2PA Content Credentials when you retrieve them through the Files API on the Claude API. Marking requires no changes to your requests or response handling.

• Like Claude Fable 5, both models require 30-day data retention and aren't available under zero data retention unless expressly authorized by Anthropic. See Model-specific data retention requirements.

• The guides for the Claude Enterprise endpoints of the Admin API (user management and spend limits), the Claude Enterprise Analytics API, and the Compliance API now show the anthropic-version header; send it on every request to these endpoints, as in the rest of the Claude API. See API versions.

We’re introducing Claude Fable 5.1 and Claude Mythos 5.1

They’re the world’s most advanced models for coding and knowledge work—and their research capabilities offer an early glimpse of how AI models will contribute to scientific progress.

Claude Fable 5.1 and Claude Mythos 5.1 are the same model, but with different levels of safeguards. Fable 5.1 is generally available, while Mythos 5.1 is available only through our trusted access programs; its safeguards are specifically designed to support work in cybersecurity and the life sciences.

Alongside its increased capabilities, Fable 5.1 takes important steps towards addressing the feedback we’ve received from customers on price, data retention, and safeguards.

Price

Fable 5.1 will cost an estimated 25% less than Fable 5 for typical workloads, wherever usage is billed by token. This is because we’re reducing our pricing on cache reads (where the model reads inputs that have already been processed and stored). For highly agentic work, the savings will often be much larger—up to approximately 45%.

Data retention

Our new system of Enterprise Frontier Safeguards (EFS) gives customers complete privacy (the same as a zero data retention policy) while still being state-of-the-art at preventing adversarial use. EFS works by storing data in cloud infrastructure controlled entirely by the customer, not Anthropic. It will be made available to enterprise customers in phases, beginning later this fall. Until EFS is available, eligible customers will be able to use Fable 5.1 with zero data retention.

Safeguards

We’ve improved our safeguards to reduce false positives (where the system flags benign content). In cybersecurity, our newest safeguards block 60% fewer false positives than before. In part, this is because Fable 5.1 can now be used to discover software vulnerabilities—though not develop exploits for them. In biology, we’ve established an access program, developed in partnership with the US government, to enable access to Claude Mythos 5.1’s advanced biology capabilities. We expect to open enrollment for scientists soon.

A new performance frontier

Claude Fable 5.1 sets a new standard on coding, knowledge work, and long-running problem-solving tasks. The charts below show that Fable 5.1 is capable of much higher performance than its predecessor, Fable 5. And when set to Low or Medium effort, Fable 5.1 achieves similar or better results than Fable 5 at a much lower cost. (Note that Fable 5.1 defaults to High effort in Claude Code, and to Medium in Claude Cowork and on Claude.ai.)

Fable 5.1 avoids shortcuts that result in poorer-quality work, and it’s smart enough to fix the root causes of software issues. For example, in testing by the investment firm Millennium, Fable 5.1 found the cause of a rare crash on their internal systems that none of their engineers (or any other model) had been able to explain after several years of trying.

Here, you can see how Fable 5.1 compares across various benchmarks:

Our early-access partners noticed these performance upgrades, and also picked up on more qualitative improvements in the model’s outputs. Here’s what they told us:

“In internal benchmarks, Claude Fable 5.1 solves more of our coding problems than Fable 5 or Opus 5, and achieves state of the art on trading intuition. While prior models became hard to follow the longer they worked, Fable 5.1 remains readable over long, multi-step tasks.”
— Craig Falls, Head of Quantitative Research, Jane Street Capital

Scientific research

We tested Claude Fable 5.1 and Claude Mythos 5.1’s scientific research capabilities across a wide range of domains. What we found—which includes the early examples we share below—adds to the evidence that AI models will soon make important contributions to scientific discovery.

Molecular design

Many modern medicines work by binding to targets within the body to block, activate, or deliver something to them. High-affinity binders are necessary for drugs to work at lower doses; designing one is the first step in the development process for many common drug modalities. To see how well Claude Mythos 5.1 could do at this task, we gave the model access to open-source protein design and folding tools and sent its designs to two external organizations for experimental validation. Mythos 5.1 proved able to design very high-affinity binders. On three targets, its binding affinities were 10 times higher than the best designs submitted to Adaptyv Bio’s protein design competitions. Its hit rate (that is, the number of designs that were viable binders) was the strongest we’ve measured to date: it reached nearly 50% across 12 targets. (Hit rates of 10-15% are typical in protein design today.)

Computational analysis and modeling

Claude Fable 5.1 trained a neural network to create a new, high-resolution elevation map of a third of the planet Venus. Its work was based on radar images taken by NASA’s Magellan mission more than 30 years ago and a map that already existed for one-fifth of the planet. Claude’s new map now reveals details down to two to three kilometers, rather than 10 to 20, and shows heights up to 25% more accurately than before.

We’re releasing this map under a Creative Commons license in advance of upcoming NASA VERITAS and ESA EnVision missions, in the hope it might help them determine which geologic features to target for future observation.

Computational biology

In computational biology, it’s common to run task-specific machine learning models on GPUs. The speed of these models is therefore a bottleneck to research progress. Mythos 5.1 provided one solution to this problem: by writing custom GPU kernels and caching their intermediate results, it sped up seven open-source deep learning models by up to 2.5 times (with identical outputs).

The benefits of such speed-ups accumulate quickly. In any given experiment, biologists might run these models thousands of times (for example, testing every possible mutation near every human gene). On analyses like these, the optimized models cut estimated GPU costs by 30 to 60%. This kind of optimization would normally take a team of performance engineers weeks, and is often unaffordable for academic labs. Mythos 5.1 was able to do it in just days, using the publicly available source code alone. We plan to open-source these optimizations soon.

As our models’ scientific capabilities improve, our investment in scientific progress is also growing. Last week, we previewed the Model Hardware Standard, which allows Claude to directly and safely operate laboratory equipment. We’ve also recently expanded our support for scientists through our AI for Science program, which provides free credits to researchers working on high-impact scientific projects, and we are offering steeply discounted usage through our new Claude Team plan for scientists.

Safety, security, and alignment

AI models’ agentic capabilities have become much more powerful over the past two years. But as we’ve documented, greater autonomy comes with new risks. Work on safety, security, and alignment needs to advance at the same pace as AI capabilities. Yesterday, we published a report describing how we are improving our own alignment and security efforts.

Prior to releasing Claude Fable 5.1 and Claude Mythos 5.1, we (and, in some cases, external researchers) subjected the models to extensive testing for risks across many areas. We describe these efforts in full in our System Card; below is a brief summary.

Chemical and biological risks

We tested the extent to which Claude Mythos 5.1 could help create chemical or biological weapons. This involved expert red-teaming, automated evaluations, and a tabletop exercise that paired PhD-level biologists with AI experts, testing whether the models could match human specialists’ performance. Mythos 5.1’s capabilities are greater than those of Mythos 5. However, our evaluations indicate that it still falls short of the next risk tier defined in our Responsible Scaling Policy. We are therefore deploying Mythos 5.1 with the same safeguards that we applied to Mythos 5, which restrict access to research biology capabilities.

Cyber risks

We ran a suite of evaluations to assess the cyber capabilities of Claude Mythos 5.1 (with cybersecurity safeguards off). Overall, the model demonstrates the strongest cyber capabilities of any model we’ve released, though it still falls within the lower category of risk in our Frontier Compliance Framework. We also performed extensive stress-testing of our cybersecurity safeguards for Fable 5.1: as well as our own dynamic evaluation of their robustness, we commissioned external testing from two organizations, along with automated testing by Gray Swan. As with Fable 5 and Opus 5, we have not found evidence of a critical-severity jailbreak for these safeguards.

Agentic safety

We ran evaluations of how Claude Mythos 5.1 responds to malicious requests and prompt injections (adversarial instructions hidden within content processed by AI models). It refused malicious agentic coding and computer use requests at a comparable rate to Mythos 5, Sonnet 5, and Opus 5, and it is our most robust model to date on an external prompt injection benchmark.

Alignment

We tested the model’s behavior through static and interactive behavioral evaluations, analyses of its internal thinking using natural language autoencoders, misalignment-related capability evaluations, a review of our training data, and analyses of our internal pilot use. We also received reports from external testing.

Our automated behavioral audit found that Claude Mythos 5.1 is better aligned across most metrics than its predecessor, Mythos 5. The model is significantly less likely than Mythos 5 to try to access resources outside of its test environment when assigned an otherwise impossible task. It is also less likely than Mythos 5 to use motivated reasoning to justify its actions (for instance, by reasoning that the situation is a simulation or evaluation), and it is less likely to ignore explicit constraints in pursuit of users’ goals. From our review of its training data, Mythos 5.1 both attempts and succeeds at reward hacking (or cheating) at a lower overall rate than Mythos 5.

Though generally our alignment evaluations showed improvements, our testing found the model can still sometimes bypass approvals and auto-mode classifiers (as we discuss in more detail in our System Card). There are also limitations to the coverage provided by our alignment assessment. Currently, our automated behavioral audit provides less visibility into very long-context work and multi-agent settings. We also have less coverage of impossible tasks (which can elicit more abnormal and misaligned behavior) than we’d like, although we’ve recently made improvements in this domain and are working hard to continue doing so.

We have also improved our safeguards so that they allow our models to be more useful without compromising on safety. We describe these changes below.

Automated safeguards for enterprises

Enterprise Frontier Safeguards (EFS) allows us to detect and respond to misuse of our models while still providing our enterprise customers the privacy of a zero data retention agreement. With EFS, customers store their data on their own cloud infrastructure, rather than on Anthropic’s systems; any human review is, by default, done by the customer themselves, rather than Anthropic. We developed EFS in close collaboration with more than 100 customers across industries like financial services, healthcare, manufacturing, telecom, law, retail, and the public sector, and with our cloud partners at Amazon Web Services, Google Cloud, and Microsoft Azure.

EFS will be supported on Claude Code, Claude Enterprise, the Claude Platform, Amazon Bedrock, Claude Platform on AWS, Google’s Agent Platform, and Microsoft Foundry. It’s rolling out in phases, starting this fall. As noted above, customers who are eligible for EFS can use Fable 5.1 (and Fable 5) with zero data retention until EFS is ready. You can read more about EFS here; to request access, please complete this form.

More precise safeguards for biology and cybersecurity

In the past few months, we’ve made progress in making our safeguards for Fable 5.1 more precise: ensuring that they’re less likely to flag benign content (like queries about medical issues or cyberdefenders using the model to make their systems safer), but still ensuring they provide robust protection against genuine threats.

As we recently shared, our latest biology safeguards for Fable 5.1 and Fable 5 fire 85% less often for benign requests related to elementary biology and medical questions (relative to those that launched with Fable 5). However, queries related to research and development in the life sciences will still be directed to our Opus models. We’re making the model’s life sciences capabilities available to professionals through an access program for Claude Mythos 5.1 that we’ve developed in partnership with the US government, which we discuss below.

With Fable 5.1, we’re updating our cybersecurity safeguards to be more precise. We’re also now allowing Fable 5.1 to be used for identifying software vulnerabilities—that is, to conduct the kind of defensive work that improves software security. As a result of these changes, Claude Code users can expect an average of around 60% fewer interventions per session from our cyber safeguards, relative to the previous safeguards on Fable 5. Our safeguards do, however, still redirect several kinds of dual-use cybersecurity tasks (tasks that might have helpful or harmful applications) to our Opus models. This includes penetration testing, exploit generation, and binary-based vulnerability scanning.

Anti-distillation mechanisms

Distillation is a method used to extract the capabilities of advanced models. It is often employed on an industrial scale, using thousands of fake accounts. Distillation is a safety risk, since the distilled capabilities can subsequently be released without adequate safeguards. Fable 5.1 comes with strengthened mechanisms to make distillation attacks harder. For example, it is no longer possible for new API accounts (those created from today onwards) to manually edit Claude’s prior context in a multi-turn conversation while preserving the transcript of Claude’s prior thinking. This closes off a common, publicly documented distillation technique, which allowed distillers to illicitly extract Claude’s thinking. We’re rolling out the change gradually, to minimize disruption: existing accounts are not currently affected by this change, though it will apply to all users with future model releases. A small number of customers’ custom integrations will then be affected. Our Help Center article explains more about this change and the adjustments that developers can make.

Trusted access for Claude Mythos 5.1

Claude Mythos 5.1 is identical to Fable 5.1, but it offers more permissive safeguards for vetted individuals and organizations whose work is affected by the cybersecurity and life sciences restrictions outlined above. It will be available through two trusted access programs:

• Cyber Verification Program: The CVP currently provides access to certain Opus and Sonnet-class models with reduced cyber safeguards for defensive security work. In the near future, this program will also include access to Claude Mythos-class models. To apply to join the CVP, click here.
• Life Sciences Verification Program: The LSVP is designed so that life sciences professionals can use Claude Mythos 5.1 with safeguards designed for professional research and development activities (while all other safeguards remain in place). In partnership with the US government, we have enrolled our first participants, and we plan to expand access to this program to the broader life science community.

In addition to these trusted access programs, Claude Security, our product that scans codebases for vulnerabilities and suggests patches for human review, is now also powered by Claude Mythos 5.1.

Compliance with the EU AI Act

In July 2026, Anthropic (along with 190 other signatories, including several other major AI model providers) signed the EU AI Act’s Code of Practice on Transparency of AI-Generated Content.

This required us to add a watermark—a numerical way of determining the likelihood that Claude was involved in writing a piece of text—to the outputs of models released after August 2, 2026. As we recently explained, this watermark is invisible to anyone who does not have the detection API. It has no practical impact on the quality or content of Claude’s outputs and contains no information about the user, their organization, or their conversations with Claude.

The Act also required us to provide a way for users to tell whether a text likely contains the watermark. We are thus rolling out a detection API in private preview. It is currently available to eligible organizations as required under EU law (such as regulators, law enforcement, media, fact-checkers, independent researchers, educational organizations, and EU civil society groups). It is also available for enterprises who are similarly obligated to verify watermarking for their own compliance with the Act. We plan to expand access to the detection API over time. You can register interest in access here.

Cost and availability

Claude Fable 5.1 is available today on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can get started with claude-fable-5-1 on the Claude API.

As mentioned above, we have reduced the price of Fable 5.1’s cache reads (where the model reuses context it has already processed) wherever usage is billed by token, such as on our API. Cache reads now cost 75% less, or $0.25 per million tokens.

This change leads to a substantial reduction in the overall cost of running the model. For typical workloads, costs are reduced by around 25% relative to Fable 5. For complex coding and highly agentic tasks, the savings could be up to around 45%. The graph below illustrates why this change makes such a big difference:

Indexed cost of running the same workloads on Fable 5 and Fable 5.1, at usage-based pricing measured at default effort over four weeks of actual usage in August 2026. Typical workload covers Fable usage across Claude Enterprise, Claude Code, and the API. Highly agentic workload covers context-heavy, tool-heavy work, where cache reads make up most of the cost.

Fable 5.1’s pricing is otherwise the same as Fable 5’s: $10 per million input tokens and $50 per million output tokens. In parallel, we’re continuing our work to bring many of the improvements of Fable 5.1 to the rest of our model family.

As discussed above, Claude Mythos 5.1 is available to vetted cyberdefenders and life scientists. Currently, it is only available to a set of US organizations, though we’re coordinating with the US government to expand access to a broader set of domestic and international partners as quickly as possible. To register interest in access to Claude Mythos 5.1 for cyberdefense through the CVP, see here.

Today we’re announcing Enterprise Frontier Safeguards (EFS), a solution that combines the privacy of zero data retention (ZDR) with state-of-the-art safeguards for detecting misuse. EFS works by storing data in cloud infrastructure controlled by the customer, not Anthropic. EFS will be rolling out to customers in phases, starting later this fall. To make the transition smooth, eligible customers will receive ZDR on Fable 5 and Fable 5.1 until EFS is ready.

We developed EFS in close collaboration with more than 100 customers in industries like financial services, healthcare, manufacturing, telecom, law, retail, and the public sector, and with our cloud partners at Amazon Web Services, Google Cloud, and Microsoft Azure.

EFS will be supported on Claude Code, Claude Enterprise, the Claude Platform, Amazon Bedrock, Claude Platform on AWS, Google’s Agent Platform, and Microsoft Foundry.

Solving the dilemma of frontier security

Mythos-class models, like Claude Fable 5.1, represent a major increase in intelligence and agentic capabilities. However, with that increase comes the potential for both misuse and autonomous misbehavior.

Over the last few months, we’ve seen substantial evidence of attempted misuse of AI models. These range from typical forms of abuse, such as fraud, to sophisticated cyberattacks, which can include agents autonomously engaging in destructive behavior. Some of these instances involve theft or misappropriation of enterprise customers’ credentials, which are difficult to detect without the ability to monitor traffic and detect abnormal behavior.

Furthermore, because the most sophisticated misuse can involve many tasks spread across multiple sessions and accounts, it is not sufficient to run automated analysis on each interaction separately and then instantaneously discard the data. Effective detection requires storing data for a meaningful period of time so that it can be correlated across time and accounts.

For this reason, we introduced 30-day data retention starting with Fable 5. This policy was not motivated by a desire to train on enterprise data: Anthropic has never trained on enterprise data without explicit permission, and never will.

The enterprises we worked with generally understood the safety and security value of data retention, but many–especially in regulated industries–found it difficult to use models with data retention. We therefore sat down with customers to design a solution that could provide the best of both worlds: the privacy of ZDR and the safety allowed by monitoring across time and accounts.

Designed with our customers

We built Enterprise Frontier Safeguards with feedback from the experts who will use it every day: security, product, compliance, and delivery teams. One of the groups we worked with was the Analysis and Resilience Center for Systemic Risk (ARC), whose members include the chief information security officers of the largest US banks, including Goldman Sachs, Morgan Stanley, Citi, Bank of America, and Wells Fargo.

We also worked with leaders at companies such as Comcast, KPMG, Mastercard, Salesforce, and Visa, to make sure the design held up across industries. Our conversations spanned a quarter of the Fortune 100, every US global systemically important bank, and virtually every regulated industry.

Here is what we heard from this wide range of customers, and what we built into EFS to address these common concerns:

On monitoring

Enterprises have long applied monitoring for insider risk, and now want help upleveling monitoring for agents. Their concerns were about Anthropic’s automated monitoring systems meeting their regulatory standards.

With EFS, customers control how data gets reviewed. When monitoring detects a pattern that needs attention, those signals are sent directly to customers so they can review what the automated systems detected.

On data storage

It’s a lot of work for enterprises to add another “trusted data vendor” for a number of reasons. They need to notify all of their customers who these vendors are and update contracts. They also have internal requirements for safely storing and auditing data, given its high level of sensitivity. Because of these concerns, we architected EFS so that customers have the ability to store data on their existing cloud infrastructure.

In EFS, customers can control their data storage and management. Customers want the ability to have their data live in infrastructure they control, under their own encryption keys, access policies, and audit logging. Activity data used for monitoring can be stored in the customer’s own cloud account (such as Amazon S3, Azure Blob Storage, or Google Cloud Storage).

On automated and human review

Even as automated review is becoming more effective, a person looking at a flag still adds value by confirming real misuse and clearing false positives. But what we heard from many customers, especially those in regulated industries, is that the person doing that review needs to be one of their own. Many operate under rules that tightly govern who may see certain information—privileged legal material, non-public information, drug-safety reports. Their teams are already trained and cleared for that work.

EFS has automated safety monitoring, no Anthropic human review required. Customers want protection against cyberattacks, and appreciate that these can be difficult to detect if they unfold across many sessions and accounts. With EFS, automated systems analyze a rolling window of traffic for signals of serious misuse, including attempts to develop offensive cyber or biological capabilities and signs of stolen or leaked credentials. Those flags go directly to the customer and their people take it from there – no human review by Anthropic employees is required.

How EFS works

These controls are designed to work the same way whether you access Claude directly from Anthropic or through a cloud partner. Customers on Amazon Web Services, Google Cloud, and Microsoft Azure will get equivalent controls, with their activity data stored in their own cloud account, in the environment they already trust. We’re also working to support third-party offerings that serve customers that are eligible for Enterprise Frontier Safeguards.

Customer-owned storage, Customer-Managed Encryption Keys, and fully automated review are each opt-in, so you enable the ones your organization needs. None of them change model behavior, API pricing, or rate limits.

Anthropic doesn’t charge for Enterprise Frontier Safeguards. If customers elect to store their data in their cloud account, their cloud provider bills them for that storage, as well as reads, writes, and data egress fees, the same way it bills any other resource.

Getting started

Enterprise Frontier Safeguards will roll out to customers in phases, with the goal of making it broadly available later this fall. To request access to Enterprise Frontier Safeguards, please complete this form.

• Added Claude Fable 5.1 (claude-fable-5-1), now the default Fable model — 1M context, $10/$50 per Mtok with $0.25/Mtok cache reads
• Added "Time format" (timeFormat) and timeZone settings: 12-hour, 24-hour, 24-hour UTC, or a strftime pattern for the turn-end clock and transcript-view timestamps
• Added a Containment Escape rule to auto mode so cloud metadata-credential fetches, egress evasion, and cross-tenant reach are no longer auto-approved unless your environment marks them expected
• Added CLAUDE_CODE_SUBAGENT_MODEL_FORCE to apply CLAUDE_CODE_SUBAGENT_MODEL (or the main model) to every subagent, ignoring per-spawn and agent-definition model overrides
• Added s in /effort to change effort for the current session only, matching /model
• Added a /doctor warning for stale sandbox mask files left by a killed session
• Added a one-time prompt in auto mode before the first file read outside the working directories, with the option to block such reads (permissions.blockReadsOutsideWorkingDirectories)
• Added support for a gateway-supplied description on discovered /model picker entries (CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY); entries without one still read "From gateway"
• Fixed settings in a .claude/ folder created after startup not being picked up until restart
• Fixed sessions dispatched from an agent view opened with ← always starting in the original session's permission mode, overriding the target directory's defaultMode and the agent's permissionMode
• Fixed keybindings.json rebinds of Ctrl+G being ignored in claude agents; its Ctrl+S / Ctrl+T are now rebindable via the new Agents context
• Fixed background sessions failing to start on macOS npm installs during a self-update, and on Windows when a stale daemon lock file pointed at a reused process id
• Fixed the working spinner stopping while a response streams behind a slash-command panel
• Fixed a background session's state.json detail repeating its own dispatch prompt after a scheduled wake-up
• Fixed claude agents keeping a background session you re-prompted buried in Completed after it finished again; Completed now orders by the latest finish
• Fixed claude --bg from a directory that was just deleted reporting "backgrounded" and leaving a crashed session row; it now prints the reason and exits 1
• Fixed Remote Control connecting mid-session re-sending the Bash tool definition, causing a prompt-cache miss
• Fixed a doubly-listed custom Authorization header overriding the configured credential on Bedrock, Mantle, Vertex, and WIF, and the Vertex setup wizard picking up a leftover Anthropic profile from ~/.config/anthropic
• Fixed Claude apps gateway sending stray host Authorization or profile headers to Foundry, Vertex, and Bedrock, and Foundry Entra ID upstreams not starting when ANTHROPIC_FOUNDRY_API_KEY is set
• Fixed a leftover Anthropic API key or auth token being sent alongside your Foundry subscription key in API-key mode
• Fixed /schedule routines whose prompt was saved without a message role and then ran with nothing to do
• Fixed claude agents not saying that a background session is waiting for you to approve a message from another session, or who sent it
• Fixed a prompt stashed with Ctrl+S inside an opened background session being lost when the session went idle or was stopped and then reopened
• Fixed telemetry (OTEL) settings pushed through server-managed settings being ignored on warm starts, including desktop-app Code sessions
• Fixed a teammate permission request being answered twice when the leader's mailbox write was briefly locked
• Fixed a phantom duplicate slash-command row rendering below the in-flight turn while a command's auto-continued response streamed
• Fixed policyHelper timeoutMs and refreshIntervalMs values above the timer maximum (2147483647) causing failures or re-runs every millisecond; they are now clamped
• Fixed the token counter freezing or crawling after switching to another subagent's transcript, and made background subagents' and teammates' counters update live while a response streams
• Fixed sandbox network hosts written with a trailing dot (example.com.): a deniedDomains entry didn't block the host inside the sandbox, and "don't ask again" for such a host kept prompting
• Fixed dismissing the Remote Control consent prompt (Esc, or n at claude remote-control) counting as consent, so the next request connected without asking
• Fixed /mcp reconnect and enable still connecting a settings-file MCP server that a managed MCP allow/deny list or strictPluginOnlyCustomization loaded after startup should block
• Fixed claude mcp remove leaving a remote server's stored OAuth credentials behind when strictPluginOnlyCustomization locks MCP to plugin-only servers
• Fixed Remote Control (claude remote-control) sessions started from the Claude app ignoring the selected model and running on the machine's default instead
• Fixed --disallowedTools and session deny rules being dropped after the first settings reload when allowManagedPermissionRulesOnly is enabled
• Fixed --resume listing a backgrounded conversation twice and --continue reopening its stalled pre-background copy; --continue now also opens finished background sessions
• Fixed fullscreen mode not letting you click ! shell command output to expand it
• Fixed background sessions left running an older Claude Code binary piling up across auto-updates instead of being retired
• Fixed claude agents --json briefly switching the terminal to raw mode and undoing another program's terminal settings on exit
• Fixed Proactive output style sessions busy-looping with filler messages and repeated log reads instead of idling while a background command or Monitor they started is still running
• Fixed subagents stopping when a response was cut off mid-stream by a computer sleep, dropped connection, or server error; they now automatically continue instead of ending with an incomplete response
• Fixed ← doing nothing in the /btw panel inside a claude agents session: it now returns to the agents list (even mid-answer), and the panel comes back when you reopen the session
• Fixed sessions with an advisor model set missing the prompt cache on background requests (compaction, /recap, prompt suggestions) and re-sending the full conversation uncached each time
• Fixed claude -p exiting about 5 seconds after its final result while a Monitor the model armed was still running; it now waits for the watch to fire or time out
• Fixed a permissions.ask rule being skipped in auto mode when the matching command ran inside a compound command or subshell, letting it run without the confirmation prompt
• Fixed plugins being able to read files outside their own directory through a declared command, agent, skill, hooks or other component path that is a symlink; such paths are now refused with an error
• Fixed /add-dir rejecting a directory inside the current working directory; it now loads that directory's skills, commands, and agents like --add-dir does at startup
• Fixed the main agent not being told when you resume a subagent you had stopped from its transcript view
• Fixed a crash when pasting ANSI-colored text (e.g. a CI log) into dialogs like /feedback
• Fixed claude mcp add/remove hanging or exhausting memory when the project's .mcp.json is a FIFO or a device-file symlink; it now fails fast with an actionable message
• Fixed unbounded memory growth when non-JSONL data is piped into claude -p --input-format stream-json; it now fails fast with a clear error
• Fixed backgrounding a turn (← or Ctrl+B) while a subagent or other tool was running occasionally making the background session treat that tool as rejected instead of re-running it
• Fixed Bash Read()/Edit() deny rules not applying to < file redirects and reader commands like tac and egrep; a deny rule on any argument or redirect target now refuses the command
• Fixed resuming or messaging a subagent whose transcript had grown past 5 MB (for example after reading many images) failing with "No transcript found"
• Fixed worktree-isolated sessions refusing Bash loops, $VAR reads, "$(…)" and heredocs that never touch git as "too complex to verify that it stays inside the worktree"
• Fixed /model and /effort showing a prompt-cache warning after rewinding a conversation back to empty
• Fixed prompt-cache misses on every turn in long screenshot-heavy sessions once images exceeded the per-request size cap
• Fixed the Edit permission prompt's diff view rendering emoji and multi-code-point characters with incorrect widths
• Fixed WebSocket MCP server connection failures being logged as "[object ErrorEvent]" instead of the underlying error
• Fixed background sessions failing to open with "Couldn't start the background service" while another Claude Code process was downloading an npm update; the start now waits for it
• Fixed background commands that detach from their shell (for example under timeout or setsid) surviving a task stop or Claude Code exit
• Fixed Claude not being told when you stop a background command from the tasks panel or a connected client
• Fixed stopping a background subagent leaving its monitors running
• Fixed sandboxed git commands in a linked worktree losing write access to the repository's common .git directory after cd into a subdirectory
• Fixed Bedrock and Bedrock Mantle requests going silent during long hidden-thinking phases on Opus 4.7 and later, which let idle timeouts cut the connection; the stream now carries progress events
• Fixed launching Claude Code after a Claude apps gateway expired or revoked your session: it now says the session ended and offers /login instead of reporting a network error
• Fixed cloud sessions losing git/GitHub credentials for the rest of the session when the session's network proxy failed to start at launch; it now retries in the background and recovers
• Fixed leftover cc-daemon-* folders in the system temp directory after an interrupted background daemon start; the cleanupPeriodDays retention sweep now removes them
• Fixed Bash permission checks auto-approving certain [[ ]] conditionals that zsh parses differently from bash; these commands now prompt for approval
• Fixed the managed-settings approval prompt showing the generic warning instead of its telemetry wording when the settings also turn detailed tracing or raw API body logging off, or trace export on
• Fixed agent-team teammates in tmux/iTerm2 panes sometimes staying open after acknowledging a shutdown request
• Fixed the keyless Console sign-in ("Sign in with your Console account") not applying your organization's server-managed settings, and /status not showing the Organization for that sign-in
• Improved rendering performance: less re-render work per turn in long conversations, streaming no longer slows down as the reply grows, and background-agent updates no longer re-render the whole screen
• Improved prompt input responsiveness by reducing per-keystroke rendering work
• Improved policy helper diagnostics — refresh failures now show in /status, declining the managed-settings dialog prints why Claude Code exited, and helper timeouts are reported as timeouts
• Improved /code-review --comment to post findings on GitLab merge requests via glab mr note instead of reporting the target as unsupported
• Improved notifications: an MCP elicitation or permission ask queued under another dialog now sends its idle desktop notification at the same delay as a visible ask
• Improved verbose/transcript output: async hook completion notices that arrive together now appear on one line instead of one line per hook
• Improved the embedded grep in native macOS/Linux builds: pathological patterns now fail fast instead of exhausting memory, and -m N with -A/-C prints correct context
• Improved the context-limit error to say when auto-compact is off and point to /config to re-enable it
• Improved /loop: consecutive wake-ups where Claude has nothing to do now fold into a single line in the terminal instead of printing each one
• Changed the sandboxed Bash tool prompt to no longer list allowed network hosts, so Claude attempts requests (and you can approve new hosts) instead of assuming unlisted hosts are blocked
• Updated the /model picker and the bundled claude-api skill to show Sonnet 5's $2/$10 per Mtok pricing as its standard list price rather than a limited-time promo
• Changed computer use on macOS so clicking the desktop, Dock, or a Finder window requires granting Finder via the access dialog, like any other app
• Changed /model, /fast, and /effort to also run immediately instead of queueing until the turn ends on Bedrock, Vertex, and Foundry and when telemetry is disabled
• Fixed claude remote-control exiting and stranding attached Remote Control sessions when the server drops its environment mid-session; it now recovers
• Fixed Remote Control sessions served by claude remote-control sometimes getting stuck after it was stopped and restarted, for Team and Enterprise members without an admin or owner role
• Changed the cross-session messaging inbox socket to close connections that send no complete line within 30 seconds; scripts posting to it should connect once their data is ready
• Improved the notice when resuming a conversation whose Remote Control is held by another terminal: it now says sessions on other machines can't be seen from, or reach, this one
• [VSCode] Improved history trimming in long sessions: older tool-activity rows are dropped first so your messages and Claude's replies stay visible
• [VSCode] Improved attribution of the extension's own usage telemetry to your organization when you are signed in with a Claude account, so its data-handling settings apply

• Fixed Bash commands failing with "task output swap refused (tasks dir moved or linked)" on some Macs
• Fixed "always allow" not saving in a project that has no .claude/settings.local.json yet
• Fixed Remote Control sessions hosted by Claude Desktop or VS Code stalling for minutes after a tool finished when the connection to claude.ai was degraded
• Fixed background task notifications with very large failure output (for example git errors on a full disk) making the conversation exceed the API request size limit

How it works

Claude for Teachers is now available to schools and districts as a free Enterprise offering. This includes premium Claude capabilities like teaching skills grounded in learning science and a direct connection to academic standards in all 50 states. Qualifying organizations that sign up by June 30, 2027 get a full year of free access.

In July, we introduced Claude for Teachers, which brought free access to premium Claude capabilities to verified U.S. K-12 educators, including teaching skills grounded in learning science and a direct connection to academic standards in all 50 states, with high-quality curricula mapped to them.

Starting today, Claude for Teachers is available to schools and districts as a free Enterprise offering. School and district leaders can now bring educators and staff into one centrally managed organization under a single set of K-12 Terms, with enterprise features like single sign on, role-based access controls, and domain claiming.

A school or district administrator gets verified, accepts the K-12 terms and student data privacy agreement, and connects their email domain and single-sign-on. From there:

• Educators and school staff get free access to Claude for Teachers. This includes Claude capabilities for K-12 educators like teaching skills, state standards and the high-quality curricula beneath them, and connections to the K-12 tools teachers already use.

• Usage limits match individual Claude for Teachers accounts, at no cost. Overage billing is turned off by default unless administrators opt in.

• The school or district administrators manage the accounts. Qualified schools and districts receive our suite of Enterprise features, like role-based access control and single sign-on. Admins add and remove staff, set policies, and see adoption across schools.

• Through domain capture, teachers who already verified on the school or district's domain move into the centrally managed account.

• The school or district holds the terms and privacy agreement. Anthropic provides school and district-level K-12 Terms and Data Processing Agreement, so one set of FERPA-aligned commitments applies across your entire organization. Claude for Teachers data is not used for model training purposes, and student information is protected by our K-12 Data Processing Agreement. You can read more about K-12 privacy in Claude for Teachers here.

New for back-to-school

We’re also adding new and improved resources for educators heading back to school, including:

• Two new teaching skills, co-developed with Learning Commons.

  • Lesson preparation: Claude helps you plan for the key moments of an upcoming lesson, so you're prepared and confident.

  • Check for understanding: Build short standards-aligned checks for understanding that surface what students are thinking, so you know what to teach next. Available for math at launch.

• Improvements across existing skills. We’ve made accessibility upgrades to student-facing materials and incorporated a lesson preparation step into our lesson planning skill.

• Updated Claude for K-12 Academy. Free AI fluency content including courses, guides, and classroom workflows for K-12 teachers, coaches, and district teams.

This is part of Anthropic's broader efforts in education, focused on supporting teachers and improving learning outcomes for students. Through our Beneficial Deployments mission, we continue to publish teaching skills as public goods for the community of developers building for teachers. This fall, we are also piloting an evaluation of Claude for Teachers in the Detroit Public Schools Community District, working closely with teachers to study the impact on educator wellbeing and practice.

Getting started

U.S. K-12 school and district leaders can verify their district and get started today. Qualifying organizations can sign up by June 30, 2027, for a full year of free access.

• Added PreModelSwitch and PostModelSwitch hook events (block, confirm, or annotate a model switch); SessionStart resume hooks now receive session staleness and the estimated re-cache cost
• Added live streaming of a foreground subagent's tool calls and results to Remote Control clients (background subagents, the default, still show status only)
• Added a Spend limit bar to /usage and a rate_limits.spend_limit status line field for developers behind a Claude apps gateway with spend limits
• Added a per-session prompt-cache line to /cost (hit ratio, misses, tokens re-cached, warm/cold) and a matching prompt_cache object for status line scripts
• Added attach, logs, stop, respawn, and rm to claude --help; the --resume message for a running background session now names the exact claude attach <id> command
• Fixed file tools (Read, Write, Edit) following a symlink swapped inside the working directory after the permission check, which could read or write outside the approved location
• Fixed plugin commands declared in a marketplace entry being able to point outside the plugin directory; such paths are now rejected with a path-traversal error
• Fixed project settings being able to enable detailed beta tracing or raw API body logging, and a lower-scope beta tracing endpoint bypassing an OTLP collector pinned by managed settings or a host app
• Fixed the Workflow tool reading (and quoting in errors) a scriptPath outside what the session may read before the permission check ran
• Fixed Grep and Glob not applying Read(...) deny rules to files reached through a symlinked search path
• Fixed conversations getting stuck on "text content blocks must be non-empty" errors after a turn where the model produced only thinking
• Fixed the first launch on a fresh install starting in default mode instead of auto mode for accounts whose startup default is auto mode
• Fixed Opus 5 requests failing with "effort … is not supported when thinking is disabled" when effort was xhigh/max and thinking was turned off; effort is now sent as high in that case
• Fixed replying to a message Claude Desktop delivered from another session: SendMessage to that session id now delivers through Claude Desktop instead of failing with "not reachable"
• Fixed TUI lag with many parallel subagents: per-second progress ticks now replace their predecessor instead of piling up in the transcript
• Fixed agent teams: a teammate's final answer not reaching the team lead — it now arrives in the idle notification instead of a content-free "available" notice
• Fixed background subagents being unable to reply to a message from an unnamed sibling or parent agent (from was the agent type, which is not an address)
• Fixed managed-settings disableAutoMode arriving mid-session not moving an already-running auto-mode session back to default mode
• Fixed a "switch to Opus 1M for 5x more context" tip that appeared even when the current Opus model already has a 1M context window
• Fixed Claude apps gateway sessions treating a stored Anthropic profile (e.g. a Console sign-in) as active: listing it in /status and retrying gateway 401s with it, though requests never use it
• Fixed cloud sessions telling Claude the model had changed when the host was only setting the session's initial model
• Fixed Remote Control reporting a failure when an organization's policy disables it; it now shows a single quiet notice instead
• Fixed /mcp reconnect on Remote Control showing a generic withheld-detail error instead of the real remedy when a server was disabled in another session
• Fixed --input-format stream-json: client-injected assistant tool calls sent without a message id were merged into the first one and their results lost, including when resuming older sessions
• Fixed session transcripts being silently overwritten when a directory change relocated a session onto an existing same-ID transcript
• Fixed background sessions and their subagents being unable to edit files inside a git worktree they created with git worktree add
• Fixed background sessions occasionally starting without any plugin skills (and staying that way) when another Claude Code process was refreshing the plugin marketplace at the same moment
• Fixed selecting text in an opened background session inside tmux over SSH: it now copies to the tmux buffer like a foreground session instead of falling back to OSC 52
• Fixed SDK and cloud sessions hanging indefinitely when an SDK MCP server's handshake acknowledgment was lost; the wait now times out after 70 seconds and marks only that server failed
• Fixed self-hosted runner leaving a stuck session's Bash tool processes running after the session was force-stopped
• Fixed /usage-credits for Team and Enterprise members whose admin set the org's usage-credit limit to $0: it now offers to ask the admin instead of saying a cap was reached
• Fixed --worktree --tmux with a merge-request number on a gitlab.com origin trying a doomed GitHub-style fetch first instead of fetching the GitLab ref directly
• Fixed Ctrl+G failing with "Emacs quit unexpectedly" in background sessions for editors that open /dev/tty, such as emacs -nw and micro
• Fixed an additionalDirectories entry containing a null byte crashing startup, or breaking /add-dir and later settings updates when it came from an SDK host, IDE, or hook; it is now skipped
• Fixed the MCP server menu's copy shortcut: it now says how the sign-in URL was copied instead of always claiming success
• Fixed italic text (such as the session recap line) rendering as highlighted blocks in GNU screen and in tmux sessions using a screen terminal type
• Fixed claude mcp add --header and claude mcp add-json help text naming the wrong transports
• Fixed claude ultrareview and /ultrareview waiting the full 30 minutes when the cloud session fails to start; they now stop early and report the reason
• Fixed Bash permission checks auto-approving commands that assign an arithmetic expression to an integer shell variable (e.g. OPTIND=1/0, RANDOM=2+2); these now prompt for approval
• Fixed backgrounding a turn (←, /background, --bg) while a subagent or other tool was running occasionally making the background session treat that tool as rejected instead of re-running it
• Fixed Bash Read()/Edit() deny rules not applying to < file redirects and reader commands like tac and egrep; a deny rule on any argument or redirect target now refuses the command
• Fixed resuming or messaging a subagent whose transcript had grown past 5 MB (for example after reading many images) failing with "No transcript found"
• Fixed worktree-isolated sessions refusing Bash loops, $VAR reads, "$(…)" and heredocs that never touch git as "too complex to verify that it stays inside the worktree"
• Fixed /model and /effort showing a prompt-cache warning after rewinding a conversation back to empty
• Fixed prompt-cache misses on every turn in long screenshot-heavy sessions once images exceeded the per-request size cap
• Fixed the Edit permission prompt's diff view rendering emoji and multi-code-point characters with incorrect widths
• Fixed WebSocket MCP server connection failures being logged as "[object ErrorEvent]" instead of the underlying error
• Fixed background sessions failing to open with "Couldn't start the background service" while another Claude Code process was downloading an npm update; the start now waits for it
• Fixed background commands that detach from their shell (for example under timeout or setsid) surviving a task stop or Claude Code exit
• Fixed Claude not being told when you stop a background command from the tasks panel or a connected client
• Fixed stopping a background subagent leaving its monitors running
• Fixed sandboxed git commands in a linked worktree losing write access to the repository's common .git directory after cd into a subdirectory
• Fixed Bedrock and Bedrock Mantle requests going silent during long hidden-thinking phases on Opus 4.7 and later, which let idle timeouts cut the connection; the stream now carries progress events
• Fixed launching Claude Code after a Claude apps gateway expired or revoked your session: it now says the session ended and offers /login instead of reporting a network error
• Fixed cloud sessions losing git/GitHub credentials for the rest of the session when the session's network proxy failed to start at launch; it now retries in the background and recovers
• Fixed leftover cc-daemon-* folders in the system temp directory after an interrupted background daemon start; the cleanupPeriodDays retention sweep now removes them
• Fixed Bash permission checks auto-approving certain [[ ]] conditionals that zsh parses differently from bash; these commands now prompt for approval
• Fixed the managed-settings approval prompt showing the generic warning instead of its telemetry wording when the settings also turn detailed tracing or raw API body logging off, or trace export on
• Fixed agent-team teammates in tmux/iTerm2 panes sometimes staying open after acknowledging a shutdown request
• Fixed the keyless Console sign-in ("Sign in with your Console account") not applying your organization's server-managed settings, and /status not showing the Organization for that sign-in
• Improved rendering performance: less re-render work per turn in long conversations, streaming no longer slows down as the reply grows, and background-agent updates no longer re-render the whole screen
• Improved prompt input responsiveness by reducing per-keystroke rendering work
• Improved policy helper diagnostics — refresh failures now show in /status, declining the managed-settings dialog prints why Claude Code exited, and helper timeouts are reported as timeouts
• Improved /code-review --comment to post findings on GitLab merge requests via glab mr note instead of reporting the target as unsupported
• Improved notifications: an MCP elicitation or permission ask queued under another dialog now sends its idle desktop notification at the same delay as a visible ask
• Improved verbose/transcript output: async hook completion notices that arrive together now appear on one line instead of one line per hook
• Improved the embedded grep in native macOS/Linux builds: pathological patterns now fail fast instead of exhausting memory, and -m N with -A/-C prints correct context
• Improved the context-limit error to say when auto-compact is off and point to /config to re-enable it
• Improved /loop: consecutive wake-ups where Claude has nothing to do now fold into a single line in the terminal instead of printing each one
• Changed the sandboxed Bash tool prompt to no longer list allowed network hosts, so Claude attempts requests (and you can approve new hosts) instead of assuming unlisted hosts are blocked
• Updated the /model picker and the bundled claude-api skill to show Sonnet 5's $2/$10 per Mtok pricing as its standard list price rather than a limited-time promo
• Changed computer use on macOS so clicking the desktop, Dock, or a Finder window requires granting Finder via the access dialog, like any other app
• Changed /model, /fast, and /effort to also run immediately instead of queueing until the turn ends on Bedrock, Vertex, and Foundry and when telemetry is disabled
• Fixed claude remote-control exiting and stranding attached Remote Control sessions when the server drops its environment mid-session; it now recovers
• Fixed Remote Control sessions served by claude remote-control sometimes getting stuck after it was stopped and restarted, for Team and Enterprise members without an admin or owner role
• Changed the cross-session messaging inbox socket to close connections that send no complete line within 30 seconds; scripts posting to it should connect once their data is ready
• Improved the notice when resuming a conversation whose Remote Control is held by another terminal: it now says sessions on other machines can't be seen from, or reach, this one
• [VSCode] Improved history trimming in long sessions: older tool-activity rows are dropped first so your messages and Claude's replies stay visible
• [VSCode] Improved attribution of the extension's own usage telemetry to your organization when you are signed in with a Claude account, so its data-handling settings apply

• Added --restricted (or CLAUDE_CODE_RESTRICTED=1): removes the built-in tools that run commands or code and WebFetch (unless named in --tools), keeps file tools inside the working directory, refuses bypassPermissions, and ignores user, project and local settings files

• Added experimental.cacheTtl ("5m" or "1h") to agent frontmatter: a per-agent prompt cache TTL used when no subagent TTL setting is configured

• Added claude self-hosted-runner --client-label <label> (or SELF_HOSTED_RUNNER_CLIENT_LABEL) to override the label the runner registers with (default: hostname)

• Added server-managed settings diagnostics: a startup warning when the settings fail to load, and a /doctor and /status line explaining a load failure or why they weren't fetched (Bedrock/Vertex/third-party provider, custom ANTHROPIC_BASE_URL)

• Added a warning in /web-setup when the GitHub CLI token lacks the workflow scope, since pushes to very large repositories can be rejected without it

• Added /usage-credits for Enterprise organizations billed through AWS Marketplace, self-serve Enterprise, and Enterprise trials, so members can request a higher usage limit from their admin

• Added cross-session messaging (SendMessage / ListAgents) between sessions on the same machine on Bedrock, Vertex, and Foundry, and when telemetry is disabled

• Fixed a prompt-cache miss (and lost extended-thinking context) roughly once an hour in long sessions, caused by tool definitions being re-rendered after an OAuth token refresh

• Fixed the ScheduleWakeup tool definition changing between a session and its --resume when the account had entered usage overage, causing a full prompt-cache miss on the resumed session's first turn

• Fixed Claude Desktop and Cowork sessions disappearing after 30 days: the transcript cleanup now keeps desktop-written sessions while they are in the app (unless org policy manages retention); the new desktopSessionCleanupPeriodDays setting caps the exemption

• Fixed being sent to the login screen when another Claude Code process held the token refresh lock while the session token had expired; the request now fails with a retryable error instead

• Windows: Fixed the claude agents list not responding to the keyboard after detaching from a session, or when launched in a terminal tab left in win32-input-mode

• Fixed the recommended Console sign-in in /login failing with an OAuth error before showing a sign-in URL on machines where it can't be used (for example when ANTHROPIC_API_KEY or an API key helper is set); it now falls back to the API-key sign-in

• Fixed model names in /model and fast-mode switch notices to render as code, so suffixes like [1m] display literally instead of as a link

• Fixed claude agents skipping the workspace trust prompt when the CI environment variable is set

• Fixed claude agents crashing on launch when the PR-status cache held a malformed entry

• Fixed agent view resurrecting a weeks-old background session after the machine was off: such a session now shows as stopped at its real end, and opening it asks before resuming its saved conversation

• Fixed agent view sometimes opening an older conversation, and dropping the typed prompt, when starting a new session

• Fixed claude agents: opening a stopped session that you already resumed in another terminal no longer starts a second process on that conversation; the row now says it is open in a terminal

• Fixed claude agents and claude rm refusing to delete a session ("has commits that are not pushed anywhere") when its worktree branch was already merged into your checked-out default branch (e.g. local main) but not yet pushed

• Fixed background sessions waiting silently when a PermissionRequest or PreToolUse hook prints an invalid answer: the claude agents row now names the hook and the schema error

• Fixed hooks silently treating a stdout {…} object that isn't valid JSON as plain text; it's now reported as a hook error with the parse message

• Fixed /mcp listing a project .mcp.json entry that declares the claude.ai connector type under the trusted "claude.ai" heading; it now appears under its real scope

• Fixed MCP servers whose headersHelper supplies the Authorization header falling into OAuth discovery on a 401 instead of re-running the helper and retrying the call as documented

• Fixed /login to a Claude apps gateway hanging when the managed-settings security approval dialog was required

• Fixed gateway model discovery (CLAUDE_CODE_ENABLE_GATEWAY_MODEL_DISCOVERY) never running when apiKeyHelper is the only credential

• Fixed claude logs leaving mouse tracking, bracketed paste and the alternate screen switched on in the terminal it was run from

• Fixed the trust dialog's list of repo permission rules showing a garbled character when a long rule was cut off in the middle of an emoji

• Fixed the permission mode indicator staying hidden behind the "Press Ctrl-C again to exit" hint when you press shift+tab right after ctrl+c

• Fixed /ultrareview and locally seeded cloud sessions uploading uncommitted edits to prod.env-style and *.tfvars files, or to editor swap, temp, and backup copies of credential files (e.g. key.pem.tmp, id_rsa.swo); they now stay on your machine

• Fixed Remote Control sessions occasionally never showing a permission prompt or the latest messages on the connected device after the CLI silently reconnected

• Fixed cloud sessions occasionally failing at startup when the container's session credentials were not yet readable

• Fixed claude remote-control rejecting its own flags (e.g. --spawn, --name) when a global flag or a wrapper-injected option precedes the subcommand

• Fixed startup warnings (e.g. "N MCP servers need authentication") rendering one column right of the rest of the transcript

• Fixed a backgrounded worktree session losing its checkout: the background session now holds the worktree's lock while it runs, so cleanup and git worktree remove leave it alone

• Fixed @-mentions of other sessions not matching names typed with non-Latin characters (for example Korean entered through an IME)

• Fixed an invalid crossSessionInbound value being silently ignored: it now warns and holds cross-session messages (user settings) or refuses them (managed settings) until fixed

• Fixed rate-limit, usage, and fast-mode messages telling you to run /usage-credits when that command isn't available for your organization (e.g. hidden with DISABLE_EXTRA_USAGE_COMMAND)

• [VSCode] Fixed a chat tab getting stuck on "No conversation found" when its session was never saved; it now starts a new conversation instead

• Improved the Workflow tool's prompt footprint: its description is now about 1k tokens instead of 5.7k, with the script-writing reference moved into a bundled workflow-authoring skill

• Improved the prompt-footer PR badge to check GitHub less often while the pull request is unchanged; a push or a gh pr command still refreshes it right away

• Improved managed settings: client-side timeout, MCP startup-mode, and stream-watchdog env vars no longer trigger the settings-approval prompt

• Improved /ultrareview <PR#> to check before launch that the GitHub account connected to your Claude account can access the repository, and to explain how to fix it, instead of failing after the cloud session starts

• Improved cross-session messaging: falls back to a private per-user /tmp directory when the default one can't be used, and the notice and /status name the directory to fix

• Changed shift+enter in the agent view dispatch input to insert a newline (matching the prompt); ctrl+enter now dispatches and attaches

• Changed /loop: self-paced dynamic mode and the no-prompt autonomous default are now always available, including on Bedrock/Vertex/Foundry

• Changed Anthropic telemetry export failures to log at debug level as [Anthropic telemetry] instead of [3P telemetry] OTEL diag error, so they are not mistaken for your OTel collector failing

• Changed cross-session messaging in Linux user namespaces: root-equivalent trust for unmapped owners is limited to canonical system directories

• Changed SendMessage from a subagent to another session: the result now notes that any reply is delivered to the parent session's conversation, not to the subagent

In Python SDK 1.2.0, TypeScript SDK 0.122.0, Go SDK 1.68.0, Java SDK 2.59.0, Ruby SDK 1.67.0, and C# SDK 12.44.0, client.beta.files and client.beta.skills no longer send the files-api-2025-04-14 and skills-2025-10-02 beta headers and return the same shapes as client.files and client.skills. With this change, client.beta.skills.delete() deletes a Skill together with all of its versions, and the beta Messages type BetaSkill (the container Skill reference) is renamed BetaContainerSkill. Requests that still send the beta headers keep receiving the beta shapes. See Migrate from files-api-2025-04-14 and Migrate from skills-2025-10-02.

• You can now create personal keys and service account keys in the Claude Console. They act as you or as a service account, with the same permissions, and stop working when the linked account is removed from an organization. This lets organization admins more easily track usage for each account, and ensure key usage is legitimate. These API keys can be scoped to a specific workspace or work on admin endpoints and across any workspace the account has access to. Workspace API keys remain supported as a legacy option. See API keys for more information.

As Claude becomes increasingly capable at scientific research, we are focusing on building products and programs to support the research community. In June, we launched Claude Science, a product that integrates the tools that researchers most commonly use, produces auditable artifacts, and provides flexible access to computing resources. We have also continued to broaden our AI for Science program, which provides free credits to researchers working on high-impact scientific projects.

Starting today, we are announcing a significant expansion of these efforts. We are opening 10,000 seats for scientists around the world to access Claude subscriptions for free and at discounted rates for one year through our new Claude team plan for scientists. Standard seats will be free and premium seats with 5x usage limits will be available for $15 per month. Over the coming months, we intend to extend this program well beyond the initial 10,000 seats.

Alongside these subscriptions, we are also expanding the scope and scale of our AI for Science program. To date, we’ve largely supported scientists using Claude for biological sciences. We are now looking to offer credits for researchers in other scientific fields as well, including those working on ambitious, compute heavy-research, such of the kind that resulted in progress on the Riemann zeta function and Claude’s work on protein design.

By helping scientists access and increase their usage of Claude through subscriptions, credits, and products like Claude Science, we aim to radically accelerate scientific discovery.

Ways to access Claude

To register for our Claude team plan for scientists, please complete the verification form here. You must be a principal investigator or equivalent at an academic or nonprofit research institution to qualify; once verified, you can add the researchers in your lab to your plan.

As your lab makes use of your allotted credits and requires more usage than standard or premium plans provide, you can apply to our AI for Science program for up to $50,000 in credits per project. Any researcher is eligible to apply.

For now, researchers working in biology and chemistry will still be limited to our Opus-class models. Claude Fable models will continue to block professional biology and drug development queries because of their potential dual-use risks. We’re working in partnership with the US government to establish an access program for life sciences professionals to use Mythos-class models for life sciences research and development. We have now enrolled our first participants, and expect to share more and increase access soon.