AI Assistants and Agents Release Notes

Release notes for AI assistants and agents you can use directly for chat, research, and work.

Get this feed:

Products (15)

Latest AI Assistants and Agents Updates

Nov 11, 2026
Date parsed from source:
Nov 11, 2026

First seen by Releasebot:
Nov 11, 2026
OpenClaw

release-publish/282ac0091707-1794364566

OpenClaw reduces repeated requests for plugin publication evidence to improve release performance.

perf(release): reduce repeated plugin publication evidence requests (…
Original source
Nov 11, 2026
Date parsed from source:
Nov 11, 2026

First seen by Releasebot:
Nov 11, 2026
OpenClaw

release-publish/3bf539dafb15-1794363903: fix(release): reject npm fanout that exceeds GitHub quota

OpenClaw fixes release quota handling by rejecting npm fanout that exceeds GitHub limits, while streamlining immutable replay test assertions without reducing coverage.

fix(release): reject npm fanout that exceeds GitHub quota

Co-authored-by: RomneyDa [email protected]

fix(release): reject npm fanout that exceeds GitHub quota

OpenClaw-Publication: 075c197a-482c-4f11-84fe-325e75228c8d

test(release): deduplicate immutable replay assertions

Keep the aggregate zero-immutable-requests invariant, which already implies the removed per-route checks. Recover seven test lines without weakening replay coverage or changing quota admission.

Co-authored-by: RomneyDa [email protected]

Publish openclaw/close-publication-quota-amplification-gap

OpenClaw-Publication: 96d8aa36-b131-4437-9656-0730f7d98ecb

Co-authored-by: RomneyDa [email protected]
Original source
All of your release notes in one feed

Join Releasebot and get updates from OpenClaw and hundreds of other software products.

Create account
Get updates with:
Nov 10, 2026
Date parsed from source:
Nov 10, 2026

First seen by Releasebot:
Nov 11, 2026
Gemini by Google

Gemini Intelligence: New features, on more devices

Gemini expands to more Android devices with new tools for custom widgets, personalized device support and task automation. The update also brings capabilities to more regions, with proactive Android Auto suggestions and Pixel Watch widgets coming soon.

Gemini Intelligence is now available on more devices, including motorola signature 27 and Pixel 10. See how expanded capabilities across these devices help you do more.

Gemini Intelligence

1 brings the best of Gemini to Android devices, helping your devices work harder for you. Earlier this year, we announced Gemini Intelligence, and this summer devices started rolling out — from the Pixel 11 to the Samsung Galaxy Z Fold8 Ultra, Z Fold8, Z Flip8 and S26 series. 2 Today, we’re introducing new features built for Gemini Intelligence, and expanding capabilities to more of Android’s most advanced devices. 3 Here’s a look at what’s new.

Gemini Intelligence on more devices, including motorola signature 27 and Pixel 10

Gemini Intelligence is coming to even more devices. Motorola is introducing its first Gemini Intelligence device, motorola signature 27, which is available for pre-order in the U.S. on November 19. And with today’s Pixel Drop, Pixel 10, Pixel 10 Pro XL, and Pixel 10 Pro Fold get the latest features for Gemini Intelligence.

Alongside this expansion, we’re bringing some of our favorite capabilities to more devices and regions, helping handle the busywork so you can focus on what matters. Rambler 4 — which turns spoken thoughts into clean, organized text in real-time — is now available on Pixel 10, Pixel 10 Pro, Pixel 10, Pro XL and Pixel 10 Pro Fold, as well as the new motorola signature 27. And task automation 5 — which lets you ask Gemini to handle logistics like ordering food, hailing rides, or booking reservations across your favorite apps — is expanding to the UK, India, and Japan.

Build custom widgets with a prompt

Create My Widget 6, which is rolling out over the next few weeks, lets you design a personalized widget with a simple description in your own words — no coding required. Whether you want a dashboard for your upcoming trip, a monthly spending tracker, or a race day countdown, Create My Widget can pull in information from Gmail, Calendar, and Search to create a custom widget just for you, giving you the information you want at a glance on your homescreen.

Get personalized tech support with Device Assistance

Device Assistance 7 helps you control, change, or fix your device by just asking your phone for help. Ask Gemini to clear up space on your device, check which apps can access your location, or mute distractions while making sure important calls still come through. Gemini deeply understands your device so it can act as your personal device expert.

Experience proactive help across a range of screens

Gemini Intelligence is designed to support you across whatever screen you’re using. Last month we launched Googlebook, the first laptop designed for Gemini Intelligence.

And if you have a Pixel phone with Gemini Intelligence, it will soon bring proactive suggestions to Android Auto — connecting the dots across your apps to surface helpful, personalized suggestions 8 when you need them. For example, if your friend texts you to ask where you are meeting for a pilates class, Gemini Intelligence on Android Auto can find your confirmation email and instantly suggest sending the address with a single tap.

Plus, in the next few weeks, Create My Widget is coming to Pixel Watch. You’ll be able to design widgets that keep the details that matter to you right on your wrist.
Original source
Nov 10, 2026
Date parsed from source:
Nov 10, 2026

First seen by Releasebot:
Nov 11, 2026
Gemini by Google

Gemini 4 Carbon is now available

Gemini releases its Carbon frontier model to developers, enterprises, researchers and eligible consumers, bringing advanced reasoning for complex coding, scientific research and cybersecurity. It’s available across Google’s API, business and consumer products, with Antigravity’s new multi-agent /swarm feature previewed for complex projects.
Developers, enterprises, scientists, and consumers can now try our latest frontier model.

Gemini 4 Carbon is available

Today, we’re making Gemini 4 Carbon available to developers, enterprises, scientists, and consumers. Carbon is our strongest model with frontier performance in cybersecurity capabilities and long-horizon agentic workflows across complex knowledge work, software engineering, and scientific reasoning.

Two weeks ago, we released Carbon to a set of cybersecurity defenders via our Fairwind Program. Following our recent launch of Gemini 4 Argon, Carbon pushes the frontier even further. Argon remains the ideal choice for the majority of software engineering and knowledge work, while Carbon is built for tasks that test the limits of a model’s intelligence like complex, multi-step coding and debugging and open-ended scientific research and discovery — using more tokens than Argon to explore thoroughly, reason deeply, and run extended verification loops.

Across independent benchmarks designed to test model performance in real workflows, Carbon pairs frontier accuracy with practical production economics. It ranks #1 on CursorBench 4.0, Perplexity’s WANDR, and Zapier’s AutomationBench at more than 2X lower cost per task compared to the next best external model.

To put Carbon’s capabilities to work on larger and complex projects, we’re previewing /swarm in Google Antigravity, turning a single prompt into a coordinated team of agents you can track in real time. Developers can assign Carbon to roles that demand the deepest reasoning, like architectural design, coordination, and auditing, while pairing it with faster, lighter models across parallel isolated workspaces and adversarial verifiers that test every change.

Gemini 4 Carbon is available via the API in Google AI Studio and Gemini for business for all paying developers and enterprises, at the same introductory price as Gemini 4 Argon ($2 per million input tokens and $10 per million output tokens). It is also rolling out to all Google AI Ultra subscribers in Google Antigravity, the Gemini app, and AI Mode in Search.

Higher-precision knowledge work, artifact generation, and data analysis

Across legal, finance, and enterprise operations, Gemini 4 Carbon combines frontier visual understanding, reasoning, and exhaustive retrieval to synthesize grounded insights across large sets of dense documents and data. It follows strict schemas and formatting to generate polished and precise spreadsheets, presentations, and reports.

Existing benchmarks don’t adequately capture how agents reason natively over dense files with charts, images, blueprints, schematics, and tables, so we partnered with Artificial Analysis to launch the new Multimodal Knowledge Agent (MKA) benchmark that tests a model’s multimodal performance in agentic workflows. Carbon ranks #2 while being more than 6X lower in average cost per task compared to the top performing model.

Using Google Antigravity’s new /swarm feature, a team of Gemini 4 Carbon agents turned historical photos and engineering plates into an interactive Three.js 3D exhibition of the Eiffel Tower's 1887–1889 construction from a single prompt.

With seven public City Council planning PDFs (206 extracted engineering drawing views) and a single prompt, a multi-agent Carbon team using Google Antigravity’s /swarm feature built an interactive 906K-triangle Three.js reconstruction of Google’s Platform 37 office across 10 overnight loops.

In Google Antigravity, Gemini 4 Carbon built an interactive multiscale textile visualizer based on an academic paper, modeling a virtual-fiber finite element framework with interactive 3D yarn meshes, 110-fiber polar tracking, and dynamic stress-strain curves with minimal prompting.

In pre-release evaluations, Carbon matched the performance of the best model tested on deep financial research on Hebbia’s E2E Financial Agentic evaluation, and ranked at the top out of 23 frontier model configurations on Balyasny Asset Management’s General Financial Intelligence benchmark, spanning legal expertise, agentic tool use and execution, coding and data engineering, and perception and context understanding.

Gemini 4 Carbon is the strongest model we’ve evaluated for deep financial research, building cleaner, better-structured Excel models and producing sharper PowerPoint decks. It is a real step forward for diligence workflows.

By Shanks Salwan, Lead Applied AI Engineer, Hebbia

Our benchmark is designed to look beyond general capability and evaluate the work financial teams actually do. Carbon stood out on tasks involving complex context and structured deliverables, especially charts, code, and tables.

By Audrey Liu, Senior AI Engineer, Balyasny Asset Management

Gemini 4 Carbon goes beyond research. It finds the call summary, the buried email or the exact pipeline figure across workplace data, then turns it into targeted analysis and next steps. In our evaluations, it led every frontier model we tested on task completion, and we're excited to bring it to Glean customers tackling complex, multi-step work.

By Thai Tran, AI Product Lead, Glean

Improved agentic coding, full stack development, and 3D and UI design

Carbon performs exceptionally well on engineering tasks that require multi-agent collaboration like multi-package migrations, cross-subsystem root-cause investigations, and complex architectural redesigns. For web app development, it runs build-and-verify loops, recovers from errors, and checks both its code and the rendered UI to produce complete, working applications. Stronger visual design, spatial reasoning, and attention to environmental detail also carry over to 3D game development and creative coding.

When we struggled to get our app looking consistent with an existing UI framework, we gave Carbon a list of requirements before closing our day. Overnight, Carbon had built a new Rust UI library from scratch and rewritten our OpenCode desktop app with working demos. It’s an incredibly brainy model and now a preferred planning model for our hardest tasks.

By Kit Langton, Software Engineer, OpenCode

Gemini 4 Carbon is by far the best frontier model for difficult cloth simulation. Its understanding of 3D, physics, and difficult (subtle) scientific decisions really make this model reliable. We spent 6 days working with it with minimal steering and it improved an existing, well tuned, soft body simulation in both speed and fidelity.

By Andrew Carr, Chief Scientist and Co-Founder, Cartwheel

Gemini 4 Carbon treats legacy software contracts with absolute respect. Across 16 real bug fixes in our 14,000-test codebase, Carbon achieved a 94% resolution rate with zero regressions.

By Baki Celtik, AI Engineer, HubX

Our game characters must listen, follow world rules, and execute multi-step plans in real time. In testing, Carbon excelled at this. It accurately followed complex player requests and held firm world boundaries, explaining why impossible actions were disallowed. That kind of judgment makes characters feel truly alive.

By Brian Tanner, Co-Founder & CEO, Artificial Agency

In our testing, Gemini 4 Carbon grounded its analysis in source content, adjusted its conclusions as evidence changed, and delivered clear recommendations in completed reviews. These capabilities are particularly valuable in highly regulated environments where decisions depend on the careful review of complex information. For example, Carbon scored 100% on public sector tasks, reinforcing our confidence in its ability to deliver accurate recommendations.

By Yashodha Bhavnani, VP of AI Products, Box

Using Google Antigravity’s /swarm feature, Carbon built an open-world 3D game where you explore Google’s Mountain View campus as a pelican riding a GBike — coordinating subagents to model 3D assets in Blender from Omni and Nano Banana references, build the browser game in Three.js with WebGL shaders, and pair procedural Web Audio sound effects with Lyria music. Try the live app.

Paired with improved front end taste and motion design skills in Google Antigravity, Carbon reimagined the Google Merchandise Store in two prompts, proposing five art direction concepts and generating all-new product imagery with Nano Banana 2.1. Try the live app and Gemini front end design skill.

Starting from a video reference generated with Gemini Omni in Google Antigravity, built a cinematic multi-shot underwater Atlantis scene. The pipeline modeled procedural ruins, volumetric caustic lighting, and water simulation in Blender before converting the full environment to Three.js for real-time browser rendering.

Using a multi-agent loop in Google Antigravity, Carbon reconstructed a playable c. 1780 Stein-Schule fortepiano with 61-key escapement physics from CAD drawings and X-ray radiographs and placed it in a 3D model of the Royal Conservatory of Music's Koerner Hall in Toronto with seat-specific spatial acoustics. Try the live app.

Frontier cybersecurity capabilities

Gemini 4 Carbon for cyber delivers our strongest cybersecurity capabilities to date, giving security teams the end-to-end intelligence they need in protecting critical software. When accessed through the Fairwind Program, Carbon leads on key benchmarks across finding (SEC-bench Pro), exploiting (ExploitGym), and patching (CWE-bench v1) vulnerabilities: inspecting software, tracing how flaws can be chained together across systems, and generating verified root-cause patches.

Early testers inside and outside of Google are already seeing strong results with Carbon for cyber in practice.

In Project Fortify, Google’s autonomous system spanning scanning, triage, and remediation, Carbon hit 100% recall on scanning and 95% precision on triage, on our internal benchmark. On Open Source code repository testing we see an over 25% relative improvement on Argon's already strong recall rate, surfacing previously undetected, real-world logic flaws. Bringing these gains across Google, one of the world’s largest users of open-source software, helps secure both our own systems and the broader ecosystem.

Google’s Cloud Vulnerability Research (CVR) team recently used Carbon to prove that two bugs in a popular open-source software were in fact exploitable vulnerabilities, comparable in severity to the recent PHP2Shell exploit. By using Carbon to build a proof of concept, CVR helped maintainers re-classify the bug as critical and swiftly issue a patch. Details will be shared after the responsible disclosure process completes.

As part of Wiz’s Scan for Good program, the team used Carbon to assess vulnerabilities in public digital infrastructure. For example, when scanning two major SaaS providers, the model uncovered 0-day vulnerabilities in under two hours compressing weeks of vulnerability research to a single afternoon. The team partnered with the impacted organizations to remediate in record time, helping to mitigate potential impact on tens of thousands of companies.

In a controlled evaluation of kernel driver vulnerability research, every model we tested found the same core memory-corruption bugs. Carbon for cyber went further, it understood the driver’s constraints, separated the findings that hold up from those that don’t, and uncovered four additional verified vulnerabilities. That is the difference between a finding and an actionable result.

By Chris Corde, Chief Product Officer, Sentinel One

Incorporating deep reasoning models like Carbon for cyber allows defenders to evaluate complex codebases and determine whether vulnerabilities are genuinely exploitable. Our early testing showed a level of precision by Carbon that could allow engineers to move past the noise and focus where human judgment creates the greatest leverage.

By John White, Dir. of Advanced Security Initiatives Group, Cisco

Gemini 4 Carbon for cyber is among the highest-scoring models we have ever tested, and across our evaluations it gets there in half the steps of its predecessor. One standout: an exploit that chains multiple hard-to-exploit vulnerabilities into a forged administrator session. Most models we've tested have never completed it. Carbon solved it consistently. This model is a step change for Gemini and a critical tool in the defender's arsenal.

By David Slater, Founder & Chief Architect, Armadin

Carbon for cyber impressed us right away. Its ability to methodically and systematically catalog large, diverse codebases and generate deep attack chains is best in class. The model's performance even at deep context windows stood out immediately in Chainguard Harness. We'll be integrating it even deeper into our autonomous vulnerability discovery and remediation systems.

By Dan Lorenc, Founder and CEO, Chainguard

In early testing, Gemini 4 Carbon for cyber performed well in detecting complex software weaknesses, maintaining the quality of its findings at a highly competitive cost per validated finding compared to other models we tested.

By Carl Windsor, CISO, Fortinet

Stronger scientific and mathematical reasoning

Gemini 4 Carbon shows strongly improved performance in reasoning across physics, math, and life sciences, synthesizing experimental data, testing hypotheses, and reasoning through multi-step domain problems. For drug toxicity research company Axiom Bio, Carbon is now the leading model in performance for translational science, finding novel mechanisms of toxicity they hadn’t found previously. Axiom is now moving their production scientific analyses and agents to both Carbon and Argon to unlock the creation of safer drugs.

Researchers across Google have been using Carbon in Google Antigravity to accelerate their work:

Code verification: Researchers built a new agentic framework for formal verification called AlphaProof Code and put it to work on a major security challenge: mathematically proving the actual code underlying the cryptographic protocol of Signal. Using Carbon alongside other models, our researchers translated security goals to mathematical theorems. AlphaProof Code agents then used Carbon to generate millions of lines of code in Lean, a computer language for writing mathematical proofs, helping them prove key security properties in weeks — a task that historically takes expert teams years. The result indicates Carbon’s potential to unlock scalable formal verification, long considered a gold-standard approach to building secure software.

Enzyme assay development: Provided with only lab notebooks and raw data, Carbon went beyond standard analysis to surface hard-to-interpret data anomalies, which it used to diagnose subtle wet lab experimental artifacts such as incomplete automated mixing in deep wells, leading to substantially improved reproducibility. Moreover Carbon improved methods to deconvolute chromatography peaks and caught trace sample carryover in research instruments, significantly accelerating the development of new assays and tightening the feedback loop between computation and the bench.

Protein structure prediction: Structure prediction models like AlphaFold 3 use Multiple Sequence Alignments (MSAs) of evolutionary relatives to help infer how amino acid chains fold. As a result they often struggle on understudied, viral, microbial, and eukaryotic proteins that have very few matches in standard databases. Working in a multi-agent setup, our researchers used Gemini 4 Carbon to autonomously discover remote homologs across petabase-scale unannotated sequence archives, environmental and viral sequence catalogs, and 3D structure databases. For example, across 7 low-homolog protein families from the CASP17 blind test set, Carbon increased the number of homologs retrieved by 40-fold on average, greatly increasing the information available to structure prediction models without need for retraining. While broader benchmarking is still underway, addressing this upstream search bottleneck potentially offers a promising path toward modeling hard-to-characterize regions of the proteome.

Four weeks ago, we opened applications to give 10,000 scientists access to Argon in Google Antigravity free of charge for six months to accelerate the pace of their research, and we’ve onboarded thousands of scientists already. Today, we’re expanding access to Carbon for current and future participants, and going forward, the program will continue providing access to our latest frontier models as they become available. Applications remain open and onboarding continues on a weekly rolling basis, subject to eligibility, capacity, and selection.

Progress in model safety and security

With each new model release, we continue to strengthen our safety and security mitigations, agentic misuse prevention, model alignment, and harm reduction proportionately with model capability.

Because Gemini 4 Carbon introduces stronger cyber capabilities we build on Argon’s strong safety and security stack to fortify robustness to misuse and prompt injection.

Mitigating frontier safety misuse risks: We’ve strengthened our misuse safeguards to help ensure stronger defensive reasoning does not translate into offensive risk. In simulated cyberattack testing environments seeded with ExploitBench and CyScenarioBench, Carbon blocked 99% of attack trajectories from our adaptive red-teaming agent that can iteratively revisit its attack plans after each block. The Carbon safeguards lowered the per-attempt breach rate to just 0.8% (down from 3.1% on Argon). Our safeguards also lowered the automated cyberattack query success rate to 2.17% in Gray Swan’s testing, compared to 9.86% on Argon. And external testing by Trajectory Labs found zero new universal jailbreaks across 132 hours and over 13,000 requests.

Promoting safe and aligned model interactions: When we announced Carbon, we showed significant improvements in alignment as measured by our monitor flag rates across the model’s lifecycle, as well as reductions in data fabrication. New results show that on our 375-task agentic honesty benchmark, it passed 92% of tasks (compared to 70% for Argon) — declining to fill in missing experimental data 98% of the time and flagging methodological flaws 68% of the time.

Protecting users from indirect prompt injection attacks: Two weeks ago, we shared that Carbon achieved a 0% prompt injection attack success rate on the Gray Swan IPI benchmark — marking the first time that milestone has been reached. At the same time, we know that prompt injection robustness remains an area for continued improvement. Our advanced automated red teaming continues to push the limits here and has shown that models could be further improved against sophisticated attacks in some distributions. It provides the data for us to continuously advance our models’ security to the next frontier.

For more details, read the full Gemini 4 Carbon Safety and Security report.

Getting started with Gemini 4 Carbon

Starting today, Carbon joins Argon in our family of frontier models across our products:

For developers: Build with the Gemini API paid tier in AI Studio and Android Studio, and if you’re a Google AI Ultra subscriber, you can try it in Google Antigravity and AI Studio.

For enterprises: Try it in Gemini for business via the API or model drop down, Google Antigravity, or CodeMender.

For everyday users: Try Carbon starting today in the Gemini app and AI Mode in Search if you’re a Google AI Ultra subscriber.

For scientists & researchers: Apply for a chance to get access to both Argon and Carbon via Google Antigravity’s highest usage tier free of charge for 6 months (onboarded on a rolling basis), subject to eligibility, capacity, and selection.

For cybersecurity defenders: Apply to join the Fairwind Program to get access to Carbon for cyber on a rolling basis.

See the model card to learn more about Gemini 4 Carbon and to start building with it, explore our updated Gemini 4 developer guide and documentation for best practices on migrating workflows to the model.
Original source
Nov 10, 2026
Date parsed from source:
Nov 10, 2026

First seen by Releasebot:
Nov 11, 2026
Gemini by Google

Introducing the Multimodal Knowledge Agent benchmark

Gemini launches the MKA benchmark for evaluating agents on visually complex, real-world tasks and makes Gemini 4 Carbon available through the Gemini API. The model ranks second in quality in the benchmark while costing six times less on average than the top-scoring model.

The Multimodal Knowledge Agent (MKA) benchmark, developed with Artificial Analysis, evaluates how agents reason across visually complex files to solve real-world knowledge work tasks.

In partnership with Artificial Analysis, we're launching MKA (Multimodal Knowledge Agent), a new benchmark built to evaluate how agents reason across visually complex files to solve real-world knowledge work tasks.

Introducing MKA

Most knowledge work benchmarks can be solved with text alone, without ever needing to look at an image or page layout. Yet, everyday professional work looks very different. Across many industries, day-to-day work is deeply visual — from cross-referencing architectural blueprints and tracing engineering schematics to synthesizing findings into polished decks, reports, and spreadsheets. We built MKA to measure how well agents can handle this visually complex work.

To ground the benchmark in real-world workflows, we partnered with domain experts across 10 verticals and 83 occupations where day-to-day work is inherently visual. Together, these specialists — from architects and biologists to underwriters and data scientists — created MKA's 200-task evaluation set to assist them in their jobs.

MKA's 10 verticals.

Each task reflects a multi-step project that takes a seasoned professional three to eight hours to complete, pairing a realistic prompt and a bundle of multimodal input files with an expert-written rubric of roughly 70 yes/no checks, weighted 75% on factual accuracy and 25% on visual design.

Three MKA tasks, from prompt and input files to golden output and rubric.

Starting today, you can explore the benchmark and evaluate your own models on Artificial Analysis. Every model runs in the same open-source agent harness, Stirrup, with the same tools, sandbox, and turn budget.

An agentic autorater built on a non-Google model grades each deliverable. It opens every file, renders pages, inspects spreadsheet formulas, and checks the work against the expert rubric. When domain experts graded model outputs themselves, they agreed with the autorater 96% of the time.

To keep results meaningful, Artificial Analysis runs the 200-task evaluation set and keeps it private so the tasks can't leak into training data or be found by the model via search. We've also released a separate set of 10 public tasks on Hugging Face, so anyone can inspect the format and test their own agents.

Our technical report covers the methodology and results in detail. It shows that frontier models work in very different ways. On the same tasks, the models we tested have extreme differences in efficiency, using between 30 and 130 turns on average to complete each task. We show how strong perception abilities are necessary to solve each task, and we also show how models try and fail to compensate when vision is removed.

Expanding visual frontiers with Gemini 4 Carbon

Gemini 4 Carbon achieves #2 in quality among the models we tested, and completes tasks at 6x lower cost on average compared to the top scoring model.

Gemini 4 Carbon is on the pareto frontier of quality and price for the MKA eval.

Getting started

Gemini 4 Carbon is available today via the Gemini API in Google AI Studio. Explore our updated Gemini 4 developer guide and documentation to get started.
Original source
Nov 10, 2026
Date parsed from source:
Nov 10, 2026

First seen by Releasebot:
Nov 11, 2026
OpenClaw

openclaw 2026.11.3

OpenClaw releases an update focused on safer, more reliable task handoffs, clearer tool failures, bounded authentication, and secure workspace connections. It also improves credential isolation, update recovery, scheduled jobs, and Gateway responsiveness.
2026.11.3

Highlights

Private task results: Keep private subagent completions out of streamed chat replies. (#188490) Thanks @fuller-stack-dev.

Reliable child-task handoffs: Allow a parent to hand off to replacement work after cancelling a child, without leaving the cancelled child's completion receipt blocking the handoff. (#178893; related #178565) Thanks @brianmcguire and @letsc4.

Visible tool failures: Report unanswered tool failures instead of silently ending a user-addressed turn as a success. (#182458; related #99645) Authored by @steipete. Thanks @rdleclerc.

Bounded authentication: Stop stalled provider authentication at the run deadline and propagate cancellation through authentication work. (#189144) Thanks @amittell.

Sandboxed image access: Keep required-sandbox image previews inside allowed folders, even in Full Access mode, while preserving explicit administrator image allowances. (#176203) Thanks @shakkernerd.

Verified workspace connections: Preserve saved node certificate pins when an empty fingerprint is supplied, and verify workspace-transfer certificates before sending credentials or data. (#125646, #190002) Thanks @wanyongstar and @avelev99.

Faster, responsive Gateway: Reduce startup module-file loading overhead and avoid stalls when abort-command detection encounters long punctuation runs. (#192217, #184433) Thanks @karkarl, @RomneyDa, @LeizhenZhang, and @obviyus.

Changes

No new configuration options or features.

Fixes

Credential isolation: Keep reused and cached plugin results scoped to their connection credentials, and keep other plugins' transport hooks on the provider's origin. (#178444, #177974, #189423) Authored by @steipete. Thanks @obviyus.

Search and memory credentials: Keep Firecrawl search and scrape keys with their scoped endpoints and embedding credentials with their destination. (#190706, #184966) Thanks @zhangguiping-xydt and @eleqtrizit.

Credential redaction: Redact inline environment credentials from configuration reads and reflected search credentials from Exa errors. (#190618, #178745) Authored by @steipete. Thanks @yetval.

Child-process environments: Prevent managed Signal daemons and bundled language servers from inheriting unrelated Gateway secrets. (#179198, #179415) Thanks @VACInc.

Authentication errors: Surface credential-read failures rather than treating them as missing credentials, and reject explicitly selected auth profiles belonging to incompatible providers. (#173970, #181597) Authored by @steipete. Thanks @vincentkoc.

Channel safety: Keep incoming Twitch IRC content out of parser error logs and require explicit request IDs when multiple Tlon approvals are pending. (#174552, #186243) Authored by @steipete. Thanks @mmaps.

Scheduled jobs: Handle queued cron storage failures without exiting the Gateway, and clear one-shot cleanup when a job becomes recurring. (#175678, #168985) Authored by @steipete. Thanks @yxshee, @wangmiao0668000666, and @obviyus.

Updates: Let profile-backed updates recognize their own maintenance lock and preserve cancellation when ledger writes are deferred. (#177792, #187998) Thanks @GrahamMcBain and @vincentkoc.

Update recovery: Recover safely when the initial package-presence check times out, requiring a fresh valid directory identity before replacement and retaining rollback to the original package. (#181906) Authored by @steipete.

Anthropic usage caps: Recognize exhausted plural usage-limit errors without spending the transient rate-limit retry budget. (#176186; related #176179) Thanks @sunlit-deng and @blue-dot-1.

Terminal sessions: Forward /reset soft to the Gateway instead of hard-resetting the session. (#172431) Thanks @yetval and @obviyus.

Git recovery: Preserve uncertain command failures during Git repair rather than treating an unconfirmed operation as successful. (#176006) Authored by @steipete.

Memory provider selection: Accept mixed-case none and auto provider values. (#186310) Authored by @steipete. Thanks @sagealice2002-dotcom.

Complete contribution record

This audited record covers the complete v2026.11.2..abf11ef history: 34 in-range PRs + 0 retained seed-only PRs = 34 unique PRs. The generation manifest also supplies direct commits as editorial input; the grouped notes above prioritize user impact.

Pull requests

PR #192217 Thanks @karkarl and @RomneyDa.

PR #172887 Authored by @steipete.

PR #184433 Thanks @LeizhenZhang and @obviyus.

PR #125646 Thanks @wanyongstar.

PR #190002 Thanks @avelev99.

PR #176203 Thanks @shakkernerd.

PR #174552 Authored by @steipete.

PR #179198 Thanks @VACInc.

PR #186243 Thanks @mmaps.

PR #190618 Thanks @yetval.

PR #175678 Thanks @yxshee, @wangmiao0668000666, and @obviyus.

PR #168985 Authored by @steipete.

PR #188490 Thanks @fuller-stack-dev.

PR #177792 Thanks @GrahamMcBain.

PR #176006 Authored by @steipete.

PR #189423 Thanks @obviyus.

PR #179415 Thanks @VACInc.

PR #190706 Thanks @zhangguiping-xydt.

PR #184966 Thanks @eleqtrizit.

PR #186310 Authored by @steipete. Thanks @sagealice2002-dotcom.

PR #187998 Thanks @vincentkoc.

PR #172431 Thanks @yetval and @obviyus.

PR #178444 Authored by @steipete.

PR #173970 Authored by @steipete.

PR #181597 Thanks @vincentkoc.

PR #195140 Thanks @RomneyDa.

PR #177974 Authored by @steipete.

PR #178745 Authored by @steipete.

PR #182458 Related #99645. Authored by @steipete. Thanks @rdleclerc.

PR #184791 Thanks @hannesrudolph.

PR #189144 Thanks @amittell.

PR #178893 Related #178565. Thanks @brianmcguire and @letsc4.

PR #181906 Authored by @steipete.

PR #176186 Related #176179. Thanks @sunlit-deng and @blue-dot-1.

Release verification

npm package: https://www.npmjs.com/package/openclaw/v/2026.11.3

registry tarball: https://registry.npmjs.org/openclaw/-/openclaw-2026.11.3.tgz

integrity: sha512-FBq/q0TCaoA2b7qfdV0q81fHhI6RBB1zAQfJQREc1u2TlK5khVgCUsAlXskLWthyVqRz5PWEj+oOmfYgK58dHg==

release SHA: 847611dcaa3bf0e2783335f961dd1fd0580e4d9f

full release CI report: https://github.com/openclaw/releases/blob/main/evidence/2026.11.3/release-evidence.md

release publish: https://github.com/openclaw/openclaw/actions/runs/42873705304

npm preflight: https://github.com/openclaw/openclaw/actions/runs/42805216493

full release validation: https://github.com/openclaw/openclaw/actions/runs/42804576583

plugin npm publish: https://github.com/openclaw/openclaw/actions/runs/42876183245

plugin ClawHub publish: no normal OIDC candidates

plugin ClawHub bootstrap: not needed

OpenClaw npm publish: https://github.com/openclaw/openclaw/actions/runs/42858513909/attempts/2

npm Telegram beta E2E: not supplied

Android APK: monitor the native qualification and publication jobs at https://github.com/openclaw/openclaw/actions/runs/42873705304

Original source
Nov 10, 2026
Date parsed from source:
Nov 10, 2026

First seen by Releasebot:
Nov 11, 2026
OpenClaw

openclaw 2026.9.37

OpenClaw ships an extended-stable gateway release focused on privacy, security, and reliability. It keeps sensitive inspection media private, improves recovery for updates and connections, and prevents slowdowns in Telegram and the Control UI.
This is a gateway-only extended-stable release, published from a retired maintenance line by explicit maintainer decision. This release is OpenClaw from the end of September 2026, plus critical security updates, reliability and performance fixes, and features like new model support. The latest version of OpenClaw at the time of this release is 2026.11.2

2026.9.37

Highlights

Private inspection stays private: paired-node photos and screenshots are no longer automatically attached to replies, and explicit media-read denials are honored when preparing reply attachments. (#174814, #184964) Thanks @eleqtrizit.

Responsive replies and pages: long zero-width Telegram table cells and malformed plugin SVG transforms no longer stall the Gateway or browser page. (#183217, #182674) Thanks @LeizhenZhang, @obviyus, and @SebTardif Authored by @steipete. Thanks @LeizhenZhang, @obviyus, and @SebTardif.

Linked accounts survive resets: sessions keep their person-linked model account through /new, /reset, and idle or daily rollover. Fixes #179432. (#179433) Thanks @masatohoshino Authored by @steipete. Thanks @masatohoshino.

Safer update recovery: interrupted snapshot writes preserve the previous rollback point, and profile-backed updates and Doctor maintenance no longer trip over their own maintenance ownership. (#189946, #177792, #187104) Thanks @kaguranananana, @GrahamMcBain, and @RomneyDa Authored by @steipete. Thanks @kaguranananana, @GrahamMcBain, and @RomneyDa.

Connections recover cleanly: channel accounts can recover after a manual stop times out, and ClickClack reconnects after silent network drops. Fixes #184852. (#177444, #184853) Thanks @masatohoshino, @YangManBOBO, and @obviyus.

Windows state access works across path aliases: equivalent Windows paths preserve the same state ownership instead of falsely reporting offline maintenance. Fixes #176163. (#176174) Thanks @tjarrau-dev Authored by @steipete. Thanks @tjarrau-dev.

Changes

Faster plugin admission: reuse native-reference validation within the same admission capture instead of repeatedly walking companion directories, while revalidating replacement captures and recovery copies. Fixes #162047. (#162254) Thanks @WG-Mojo Authored by @steipete. Thanks @WG-Mojo.

Fixes

Private diagnostics: omit suppressed reply text and background command contents from verbose diagnostics while retaining status metadata. Thanks @steipete.

Media boundaries: keep paired-node inspection photos and screenshots private by default, and honor explicit read denials for reply attachments without blocking authorized managed outputs. (#174814, #184964) Thanks @eleqtrizit.

Restrictive group policies: Doctor preserves explicitly empty group allowlists, inherited settings, and disabled group policies instead of populating fallback senders. Fixes #173826. (#174787) Thanks @lockhartheavyindustries Authored by @steipete. Thanks @lockhartheavyindustries.

Discovery output: escape terminal control characters in untrusted Gateway discovery fields. (#183355) Thanks @Alix-007.

State and write outcomes: recognize committed SQLite writes whose worker result cannot be serialized, avoiding unsafe replay, and preserve Windows ownership across equivalent path spellings. Fixes #176163. (#177321, #176174) Thanks @tjarrau-dev Authored by @steipete. Thanks @tjarrau-dev.

Session accounts: retain linked model-account selections and their provenance through explicit resets and automatic rollover. Fixes #179432. (#179433) Thanks @masatohoshino Authored by @steipete. Thanks @masatohoshino.

Channel recovery: recover channel accounts after timed-out manual stops and restore ClickClack message reception after silent network drops. Fixes #184852. (#177444, #184853) Thanks @masatohoshino, @YangManBOBO, and @obviyus.

Local tool connections: replace unexpectedly closed MCP loopback listeners and retain their cleanup work for shutdown rather than leaving later tool calls permanently disconnected. Fixes #139340. (#183352) Thanks @SC-10749 and @eli-pavlov Authored by @steipete. Thanks @eli-pavlov and @SC-10749.

Telegram and Control UI: bound the processing of long zero-width table cells and malformed SVG transform strings without changing valid table or icon rendering. (#183217, #182674) Thanks @LeizhenZhang, @obviyus, and @SebTardif Authored by @steipete. Thanks @LeizhenZhang, @obviyus, and @SebTardif.

Command results and temporary files: preserve a completed command's exit result when detached-descendant cleanup is uncertain, retain a visible cleanup warning, and remove temporary MCP configuration directories when writing fails. Fixes #179555. (#185636, #179603) Thanks @xiaopan-xydt and @Yolo521-bot Authored by @steipete. Thanks @Yolo521-bot and @xiaopan-xydt.

Gateway credentials: keep local password SecretRefs active when a separate remote token is configured, and retain a saved node TLS pin when the fingerprint option is empty or whitespace-only. (#128175, #125646) Thanks @TheAngryPit, @CassieMei, @obviyus, @RomneyDa, and @wanyongstar.

Update and restart reliability: preserve rollback snapshots after interrupted writes, release completed maintenance before reopening requester policy, confirm predecessor stop receipts during Doctor maintenance, and prepare configuration restarts outside the restart-signal fence. (#189946, #177792, #187104, #172777) Thanks @kaguranananana, @GrahamMcBain, and @RomneyDa Authored by @steipete. Thanks @RomneyDa, @kaguranananana, and @GrahamMcBain.

Memory jobs and imports: leave stopped or failure-disabled dreaming jobs disabled, and refresh imported ChatGPT transcript examples without overwriting real Notes sections. Fixes #180655, #193161. (#180694, #193173) Thanks @Yun-0000, @lisandromachado, @KhanCold, and @obviyus Authored by @steipete. Thanks @lisandromachado, @KhanCold, @Yun-0000, and @obviyus.

Release artifact verification: qualify the candidate's exact plugin artifacts, route validation through the admitted verifier, and accept successful admission jobs that omit step details while preserving run, attempt, artifact, and authority checks. (#164464, #173437, #188597) Thanks @RomneyDa.

Complete contribution record

This audited record covers the complete v2026.9.36..2f7023b history: 28 in-range PRs + 0 retained seed-only PRs = 28 unique PRs. The generation manifest also supplies direct commits as editorial input; the grouped notes above prioritize user impact.

Shipped baseline exclusions: v2026.9.36 (1 PRs: #158414).

Pull requests

PR #162254 Related #162047. Authored by @steipete. Thanks @WG-Mojo.

PR #183355 Thanks @Alix-007.

PR #183217 Thanks @LeizhenZhang and @obviyus.

PR #182674 Authored by @steipete. Thanks @SebTardif.

PR #128175 Related #128171. Thanks @TheAngryPit, @CassieMei, @obviyus, and @RomneyDa.

PR #188597 Thanks @RomneyDa.

PR #187104 Thanks @RomneyDa.

PR #193173 Related #193161. Thanks @KhanCold and @obviyus.

PR #185636 Authored by @steipete.

PR #179603 Related #179555. Authored by @steipete. Thanks @xiaopan-xydt and @Yolo521-bot.

PR #179433 Related #179432. Authored by @steipete. Thanks @masatohoshino.

PR #167356 Thanks @vincentkoc.

PR #173437 Thanks @RomneyDa.

PR #180694 Related #180655. Authored by @steipete. Thanks @Yun-0000 and @lisandromachado.

PR #177792 Thanks @GrahamMcBain.

PR #184853 Related #184852. Thanks @masatohoshino.

PR #164464 Thanks @RomneyDa.

PR #183352 Related #139340. Authored by @steipete. Thanks @SC-10749 and @eli-pavlov.

PR #189946 Authored by @steipete. Thanks @kaguranananana.

PR #174787 Related #173826. Authored by @steipete. Thanks @lockhartheavyindustries.

PR #125646 Thanks @wanyongstar.

PR #174814 Thanks @eleqtrizit.

PR #177321 Authored by @steipete.

PR #172777 Thanks @RomneyDa.

PR #184964 Thanks @eleqtrizit.

PR #172669 Authored by @steipete.

PR #177444 Thanks @YangManBOBO and @obviyus.

PR #176174 Related #176163. Authored by @steipete. Thanks @tjarrau-dev.

Release verification

npm package: https://www.npmjs.com/package/openclaw/v/2026.9.37

registry tarball: https://registry.npmjs.org/openclaw/-/openclaw-2026.9.37.tgz

integrity: sha512-ETRzzF0mllvQrzKJp8meZ2iTVytTaPDfEc5Oh7qh8FK8c5bcJWcKH+sYxyruAqkEzb8CaB6oTlc38eZf1ufRVw==

release SHA: 81da850122080b7187816ab70f145f3bc9e5efff

full release CI report: https://github.com/openclaw/releases/blob/main/evidence/2026.9.37/release-evidence.md

release publish: https://github.com/openclaw/openclaw/actions/runs/42866233501

npm preflight: https://github.com/openclaw/openclaw/actions/runs/42782571264

full release validation: https://github.com/openclaw/openclaw/actions/runs/42782249510

plugin npm publish: https://github.com/openclaw/openclaw/actions/runs/42883284005

plugin ClawHub publish: no normal OIDC candidates

plugin ClawHub bootstrap: not needed

OpenClaw npm publish: https://github.com/openclaw/openclaw/actions/runs/42885900163

npm Telegram beta E2E: not supplied

Original source
Nov 10, 2026
Date parsed from source:
Nov 10, 2026

First seen by Releasebot:
Nov 11, 2026
OpenClaw

release-publish/358722b1a46c-1794333870: fix(release): explain rejected child evidence reads

OpenClaw improves release diagnostics for child evidence reads, preserving GitHub transport details and historical checks while explaining allowed failure reasons without mistaking every HTTP 403 for rate limiting.

fix(release): classify child evidence read failures

Preserve the original gh transport and complete historical evidence checks while reporting allowlisted failure reasons from the original response. Do not infer rate limiting from HTTP 403 alone.

Co-authored-by: RomneyDa [email protected]

fix(release): explain rejected child evidence reads

OpenClaw-Publication: 76f4b6e2-7537-4994-98b6-ab5bbe7cbfaa

test(release): use Node for child evidence fixtures

Keep extensionless fake gh commands on the canonical Node runtime with explicit CommonJS scope when the test runner is Bun. Preserve exact failure and request assertions; format the test with the pinned formatter.

Co-authored-by: RomneyDa [email protected]

Publish openclaw/child-evidence-diagnostic-pr

OpenClaw-Publication: e2def52a-2270-41be-9294-ddfc6f34db08

Publish openclaw/child-evidence-diagnostic-pr

OpenClaw-Publication: e88c5d4e-109f-49ee-ade2-053c0be33b5d

Co-authored-by: RomneyDa [email protected]
Original source
Nov 10, 2026
Date parsed from source:
Nov 10, 2026

First seen by Releasebot:
Nov 11, 2026
OpenClaw

release-publish/179529ec9dd8-1794288891: fix: prevent prepared npm authorization CLI deadlock

OpenClaw fixes a deadlock in its prepared npm authorization CLI by resolving an import-cycle issue, preserving error reporting and exit status, and strengthening subprocess validation.

fix: prevent prepared npm authorization CLI deadlock

Let module evaluation finish before the async CLI operation returns through the validator import cycle. Preserve caught error reporting and exit status; cover the real authorize subprocess through post-import validation.

Co-authored-by: RomneyDa [email protected]

OpenClaw-Publication: ac24e226-09f0-48a2-85d0-5128eefd4d39

fix: satisfy prepared npm authorization lint contracts

Type the caught rejection as unknown and await the two successful authority checks. Preserve the reviewed CLI import-cycle repair and validation guards.

Co-authored-by: RomneyDa [email protected]

OpenClaw-Publication: da54a2f7-3e80-45f3-b503-95f6b0839d0c

Co-authored-by: RomneyDa [email protected]
Original source
Nov 10, 2026
Date parsed from source:
Nov 10, 2026

First seen by Releasebot:
Nov 11, 2026
OpenClaw

release-publish/09723cdcad18-1794299463

OpenClaw fixes repeated payload downloads during receipt recovery.

fix(release): avoid repeated payload downloads during receipt recover…
Original source

AI Assistants and Agents Release Notes

Products (15)

Latest AI Assistants and Agents Updates

release-publish/282ac0091707-1794364566

release-publish/3bf539dafb15-1794363903: fix(release): reject npm fanout that exceeds GitHub quota

fix(release): reject npm fanout that exceeds GitHub quota

fix(release): reject npm fanout that exceeds GitHub quota

test(release): deduplicate immutable replay assertions

Publish openclaw/close-publication-quota-amplification-gap

Gemini Intelligence: New features, on more devices

Gemini Intelligence

Gemini Intelligence on more devices, including motorola signature 27 and Pixel 10

Build custom widgets with a prompt

Get personalized tech support with Device Assistance

Experience proactive help across a range of screens

Gemini 4 Carbon is now available

Gemini 4 Carbon is available

Higher-precision knowledge work, artifact generation, and data analysis

Improved agentic coding, full stack development, and 3D and UI design

Frontier cybersecurity capabilities

Stronger scientific and mathematical reasoning

Progress in model safety and security

Getting started with Gemini 4 Carbon

Introducing the Multimodal Knowledge Agent benchmark

Introducing MKA

Expanding visual frontiers with Gemini 4 Carbon

Getting started

openclaw 2026.11.3

Highlights

Changes

Fixes

Complete contribution record

Pull requests

Release verification

openclaw 2026.9.37

2026.9.37

Highlights

Changes

Fixes

Complete contribution record

Pull requests

Release verification

release-publish/358722b1a46c-1794333870: fix(release): explain rejected child evidence reads

fix(release): classify child evidence read failures

fix(release): explain rejected child evidence reads

test(release): use Node for child evidence fixtures

Publish openclaw/child-evidence-diagnostic-pr

Publish openclaw/child-evidence-diagnostic-pr

release-publish/179529ec9dd8-1794288891: fix: prevent prepared npm authorization CLI deadlock

fix: prevent prepared npm authorization CLI deadlock

fix: satisfy prepared npm authorization lint contracts

release-publish/09723cdcad18-1794299463