Claude Code model discovery, property filter operators, session continuity, and steadier providers

NEW FEATURES

The model picker shows only the Claude Code models your machine can actually run.

Models you can't use are called out up front, not after a run fails.

Fable 5.1 is available, with its pricing already in place.

Filter Issues by what a custom text, number, date, or URL property contains.

Number and date properties can also be filtered by range.

See which agent runs are still active from the command line, across an Issue and its sub-Issues.

Custom status colors show up on Issue links, hover cards, and editor suggestions in chat.

IMPROVEMENTS

Recovering a handed-off task that failed no longer slows down everything else running.

Japanese and Korean read consistently across the product and the docs.

Shortened text keeps emoji intact instead of cutting them in half.

BUG FIXES

Editing an Issue description saves cleanly, with no false conflict prompts.

A task that fails while preparing says so, instead of blaming the model or the agent.

Task logs show the full task ID, so parallel runs are easy to tell apart.

A replacement task waits for the previous one to release the working folder.

Every task in the same session keeps working on the same branch.

Repositories you already checked out follow your Co-authored-by setting right away.

Kimi reports provider errors instead of finishing with an empty reply.

Pi sessions run on Windows.

Codex can be set to Standard speed on purpose.

Background cleanup only removes a working folder that belongs to the task.

Nothing is quietly dropped from saved data when unusual characters are cleaned up.

Long Markdown lists on iOS no longer overlap while you scroll.

Images in task details load without flickering.

Dates and amounts follow the language you picked.

The daemon controls in the desktop app are fully translated.

Faster Issue lists, steadier long runs, and more complete localization

NEW FEATURES

• Use Huawei Cloud CodeArts as a native agent runtime.
• Pick its models, pick up past sessions, and bring your MCP servers and local Skills.
• Install Multica natively on iPad, in any screen orientation.
• WeCom bot replies still arrive when you self-host Multica across several servers.
• See how many WeCom replies never made it out, and why.

IMPROVEMENTS

• Issue lists open faster in large workspaces.
• Agents start faster when a lot of Skills are available to them.
• Onboarding, Squads, the editor, statuses, and priorities read naturally in every language.
• Agent output shows full file paths again, with nothing hidden by mistake.
• Self-hosted servers do less background work against the database.
• A self-hosted server keeps answering even when a connection stalls.

BUG FIXES

• Chat stays on the newest reply, with no jumping or flickering while it streams.
• The Analytics leaderboard scrolls sideways in narrow windows instead of being cut off.
• Codex sessions start reliably instead of giving up during setup.
• Local Skills load even when the daemon on your machine has gone idle.
• The desktop app brings back its daemon if it stops unexpectedly.
• The Issue header on phones is compact again, with avatars and tap targets intact.
• Self-hosted upgrades from older versions finish without manual repair.
• Pi and Oh My Pi keep their session when the working folder changes.
• Deleting a workspace now says exactly what will be removed.
• Long runs and queued tasks are no longer cancelled just for waiting.

Custom property filters, Skill imports, and MCP for Oh My Pi

NEW FEATURES

Filter Issues by custom text, number, date, and URL properties.

Or find just the Issues where a custom property has no value yet.

These filters show up in the filter menu, in facets, and in saved views.

Create a Skill by importing a local folder, a .skill file, or a .zip.

Preview what's inside before you import, and catch name conflicts up front.

Oh My Pi agents can now use MCP servers you manage on the agent page.

IMPROVEMENTS

Runtimes are only reclaimed when it's safe, so nothing disappears mid-run.

A runtime that has been offline for a long time now says exactly that.

Long agent steps get up to two hours of idle time before they stop.

The CLI now spells out the page size and paging options for listing Issues.

BUG FIXES

Links you copy or share from the desktop app open in any browser, on any device.

Repository checkouts keep working even if the folder you started from is gone.

Runs on long-context models are priced correctly, so usage is no longer undercounted.

Custom Issue statuses without Latin letters show the right name everywhere.

When an edit conflicts, each choice sits right under the version it applies to.

Lark replies no longer stall after a credential expires.

Quick Create restores your draft at the right height again.

OpenClaw runs stop completely when they time out, leaving no processes behind.

Setting up OpenClaw no longer fails when it already had everything it needed.

Session controls in every chat channel, agent conversation starters, and Inbox filters

NEW FEATURES

Start a fresh session with /new, or clear the current one with /clear.

Both work in DingTalk, Lark, Slack, Telegram, and WeCom.

Give an agent up to three conversation starters, shown in every empty chat.

Tapping a starter fills the composer, so you can edit it before sending.

Filter Inbox notifications by who started them, or show only unread ones.

IMPROVEMENTS

Run costs for Qwen, Kimi, and Ark models are now accurate.

The Autopilot list shows subscribers and the next scheduled run.

A single Cloud address turns on cloud runtimes, billing, and seat management together.

Local task folders now have readable names, so your files are easy to find.

BUG FIXES

Autopilots can hand work to your private agents and Squads again.

An Autopilot keeps its Project when you edit its other settings.

Chat follows the newest output as it streams, and scrolling up no longer snaps you back.

Private runtimes are no longer visible to other members.

Billing reflects your Pro plan right after you subscribe.

Codex runs no longer end early, so nothing is missed or counted twice.

Autopilot commands no longer accept a priority option that did nothing.

Issue boards stay fast even when descriptions are very long.

Reading large Skills is reliable and no longer stops midway.

Sub-Issues from comments, and custom Issue statuses for every workspace

NEW FEATURES

Turn any comment into a sub-Issue with one click.

New sub-Issues carry the original discussion, so whoever picks them up starts with full context.

Custom Issue statuses are now open to every workspace — create your own right away.

IMPROVEMENTS

Checkout pre-fills your account email automatically.

Failed repository checkouts now say exactly what went wrong and how to fix it.

Self-hosting guides now cover health checks, single-domain setups, and choosing a release.

BUG FIXES

MiniMax Code no longer fails intermittently on startup.

Members who leave a workspace no longer get its Autopilot notifications.

Automatic recovery replies for failed tasks now land in the original comment thread.

Comment timestamps no longer show the wrong timezone.

Agents locate Issues by full ID faster and more reliably.

Cancelling a task now truly stops the agent — nothing keeps running in the background.

Cancelled subscription seats no longer appear in billing.

When a task is cancelled, the CLI no longer asks you to sign in again.

Inbox filters, the ZeroClaw runtime, and localized failure messages

NEW FEATURES

• Filter Inbox notifications by Issue status and priority.
• Use ZeroClaw as a native agent runtime.
• Self-hosted teams can send daemon traffic to a separate server.
• Self-hosted teams can set how long queued tasks remain available.

IMPROVEMENTS

• Task failure messages now appear in your chosen language.
• Grok shows each model's supported reasoning levels.
• Skill download failures now show how much data arrived.
• Use a keyboard shortcut to show or hide the right sidebar.
• The iOS app now has a cleaner flat icon.
• Self-hosting guides now state the right PostgreSQL requirements.

BUG FIXES

• Windows Qwen and Pi runs now preserve complete prompts.
• Antigravity responses keep their original line breaks.
• Lark post mentions reach the intended people.
• Agent and Issue activity stays correctly authorized and ordered.
• New chats and project work stay in the right workspace context.
• Self-hosted health checks and concurrent local work stay reliable.
• Hermes, Kimi, and Pi now handle runtime failures more reliably.
• Purchase quotes and archived Inbox counts now recover correctly.

DingTalk groups, Project views, and smoother teamwork

NEW FEATURES

Mention a DingTalk bot in a group and that group appears under its agent.

Group Board and Table views by Project to keep related Issues together.

IMPROVEMENTS

Long-running agents stay more responsive while they work.

Self-hosted servers recover automatically after a brief database outage.

Telegram now greets you and replies in English.

BUG FIXES

Running a task again no longer stops work already in progress.

Tasks running at the same time no longer overwrite one another’s workspace.

OpenClaw, Pi, and OpenCode work with the custom providers you configure.

Changing workspaces no longer interrupts live updates or @mentions.

Autopilot webhook secrets stay out of normal CLI output.

Chat with your agents without waiting for unrelated coding work to finish.

Run confirmation before leaving backlog, Webhook event filters, and Windows build fixes

IMPROVEMENTS

The web Help menu now makes the desktop app easier to find.

You can confirm an agent run before moving an assigned Issue out of backlog.

Webhook event filters are easier to add and remove.

The / picker now ranks exact skill matches first.

BUG FIXES

Windows command-line builds now create correctly named executables.

Agents can still load skills after a brief connection drop.

Run now no longer exposes internal error details.

Agent updates stay responsive in busy workspaces.

Safer Issue editing, empty-property filters, and clearer agent identity

NEW FEATURES

You can see which agent is handling each chat session.

Workspace admins can see linked DingTalk identities.

You can filter Issues by custom properties that have no value.

When an Issue or comment changes while you edit it, you can compare both versions before saving.

IMPROVEMENTS

The command palette now includes every workspace page.

Chat sessions now resume faster.

Issue activity times now reflect meaningful updates.

BUG FIXES

Private agents can only be invoked by their owners.

Agent workspaces stay isolated when tasks run at the same time.

Provider command logs no longer expose credentials or prompts.

Posting a comment no longer reports a false send failure.

A dependency security vulnerability has been fixed.

MiniMax Code, cleaner run transcripts, and Hermes agent fixes

NEW FEATURES

You can now run agents with MiniMax Code.

Press Cmd/Ctrl+, to open Settings in a new desktop tab.

IMPROVEMENTS

Hermes agent runs now show tools while they are working.

Agent transcripts are easier to scan without repeated names.

BUG FIXES

Tasks keep running through brief runtime connection drops.

Daily and weekly cost charts now include cache-read costs.

Files and audio sent directly in Feishu now reach your agent.

Failed delegated work now returns to the coordinating agent.

Downloads now show the newest complete release.

Member custom fields, bulk skill updates, and a clearer desktop update prompt

NEW FEATURES

You can now put workspace members in a custom field, one or several.

Selected skills can now be updated from their source all at once.

IMPROVEMENTS

The desktop update prompt now links to what changed in that version.

The command line now accepts a member's email wherever an assignee is expected.

BUG FIXES

Opening a new tab in the desktop app no longer loses your workspace.

Member filters now show names and avatars instead of a bare count.

Workspace MCP servers, share-link invites, and a rebuilt transcript

NEW FEATURES

You can now set up MCP servers once for a workspace and hand them to individual agents.

You can now create a share link that lets someone join your workspace directly.

The transcript now reads as steps, with a two-lane timeline and an outcome summary.

Files sent in Slack now arrive in the conversation as attachments.

You can now add Multica to your phone's home screen and open it like an app.

Browser tabs are now named after the workspace page you have open.

Imported skills now show where they came from.

A mention of the Issue you are reading now reads as "This issue".

IMPROVEMENTS

An agent whose CLI cannot run now goes offline and tells you the command that fixes it.

Typing an Issue identifier now brings up the mention far faster.

Channel messages no longer duplicate or drop out across multiple servers.

Workspace invitations can no longer be sent in floods.

The editor's floating toolbar is now readable by screen readers.

BUG FIXES

Reopening the desktop app after a failed start no longer makes you sign in again.

Deleting your last workspace no longer leaves the desktop app blank.

A Pi task that ends its turn on an error now fails instead of hanging.

The Antigravity model list now loads.

HTML attachments now return to where you were reading after a tab switch.

Custom runtimes now receive their fixed arguments in the order you wrote them.

Cancelling a drag mid-way no longer locks the board.

Quick capture now works on mobile web.

Page headers no longer show a second sidebar toggle.

Creating an agent no longer flashes an error.

Hovering a bar in the run timeline no longer makes it shake.

A self-hosted upgrade that gets interrupted now carries on safely when you run it again.

DeepSeek Harness runtime, board drag-to-pan, and an Inbox archive shortcut

NEW FEATURES

You can now run your agents on DeepSeek Harness.

Dragging blank space on the board now pans it sideways.

Pressing E in the Inbox now archives the open notification.

Self-hosted setups can now keep task workspaces on any disk you pick.

IMPROVEMENTS

Page titles and toolbars now line up on the same left edge everywhere.

You can now give Codex longer to get going when its first turn is slow.

Assigning an Issue or changing its status can now skip starting a run.

BUG FIXES

An agent CLI that cannot run now tells you how to repair it.

Someone else's private runtime can no longer be used from the API or CLI.

Leftover task files no longer break every multica command in a folder.

Agents installed through Volta or Vite Plus are now recognised.

The gap between the last reply and the composer is back.

In dark mode the active tab no longer prints dark squares, and hover corners stay round.

An agent no longer starts a second run for an Issue it is already on.