AI agent releases: Aug 3 – Aug 9, 2026
230 stable releases from 89 tracked agents and frameworks in the week of Aug 3, 2026, newest first, with the headline changes from each set of release notes. Our lists keep each project's latest releases, so a very active project's older releases can be missing.
Aug 9, 2026
- feat(agents): add verification ownership budgets
- chore: fix pre-existing lint warnings
- fix(board-injection): gate synthetic part log on task-status shape
- Fix: the 0.9.37 callback-body fix no longer lets a local declared in one callback suppress a call in a sibling callback
- Fix: Kotlin calls in a property initializer are now collected
- Fix: Swift receiver-type inference now handles @Environment(Store.self) properties and factory-initialised bindings
- GenOffice-0.6.13-arm64.dmg
- GenOfficeSetup-v0.6.13.exe
- genoffice_0.6.13_amd64.deb — Debian / Ubuntu (recommended): sudo apt install ./genoffice_0.6.13_amd64.deb
- Fireworks: track 30-day rated billing spend with an API key and account slug. Thanks @x0mh0x!
- PTY probes: abort output-overflow children through a bounded process-group kill path, avoiding minute-long cleanup stalls under load (refs #2792).
- PTY runner: drain short-lived commands to terminal closure after exit, preventing false empty probe results when output is still buffered.
- Unassign keyboard shortcuts and see the effective bindings in shortcut help
- Choose a Pure black theme
- Running agents use a clearer animated status ring (c0daf8e
- Desktop session panels can expand the focused session for more working space
- Switching models preserves the active catalog profile
- Tab focus changes avoid unnecessary full terminal repaints
- Sidebar sessions grouped by live activity and date. The workspace sidebar now distinguishes live work (running, waiting-on-user, waiting-permission, waiting-plan-approval) from recently completed…
- Context compaction activity in the transcript. Automatic context compaction is now visible as a branch-scoped activity row with spinner, completed, cancelled, and failed states.
- Keyboard-driven search navigation. Composer mention results now support Tab selection, and platform search shortcuts navigate Settings filter fields.
- Closes #217, reported
- Both agents read task_plan.md, both write it back, and the later write discards the earlier one's phases.
- Attestation was the nearest existing mechanism and did not cover it. It is opt-in in legacy mode, it compares against a baseline a human approved once rather than against what the hooks last…
- create record on save instead of on Add click (8b6b978)
- commit: 7658556e653a60a6a343278f070695bbc88da2df
- source tag: release2026-08-09_10-47-58
- regions: cn global
- docs(skills): clearer description
- fix(tui): make the mount prompt clickable, and skip the mount instead of abandoning the scan
- test(tui): correct the nudge the internal-turn test asserts
- Fixed an issue where unsigned macOS uv archives were incorrectly included in the notarized app bundle, ensuring the app passes macOS notarization correctly
- fix(desktop): keep unsigned macOS uv archives out of notarized app
- macOS Apple Silicon (arm64)
Aug 8, 2026
- Split runtime context from coding agent and add project ID
- Update snapshot and changelog for v1.15.13
- Fix: TypeScript member calls no longer fabricate a high-confidence calls edge by matching a receiver type by name alone
- Fix: TypeScript/JavaScript calls inside a callback body passed to another call (for example export const handler = wrapper(async (req) => { helper() })) are no longer dropped
- Fix: Kotlin imports, fully-qualified calls, and one-line type bodies
- Extract from multiple sources in one run — each tracked with its own source_id and completeness record. Op-union (keep/set/remove) lets the same criterion be built from contributions across sources.
- Queued drain: extract --queued — pre-drawn work items are claimed from a durable queue and processed in budget-aware batches (--budget n; default 10 per composite root).
- Agent failures inside a queued batch no longer abort the whole batch — a failed claim is recorded as AGENT_ERROR and the batch continues. Early exits (including pause) release claims cleanly.
- ACP clients now receive richer protocol support and can use the full expected tool surface
- The desktop app now offers a visible new-session action and consistent hover feedback
- Cloud relay and session settings now use neutral names while continuing to accept legacy configuration keys
- feat(litellm): pass-through cache\_control\_injection\_points for Anthropic prompt caching
- feat: add GLM-5.2, Kimi K3, Qwen3.8 and MiMo v2.5 model support
- feat: add opt-in agent run details section (model, tokens, time, AI calls)
- Smarter dataset indexing for better search results
- More reliable and faster ingestion of large files
- Improved developer SDK and clearer error messages
- feat(desktop): unify add agent flows (02f640bc4559c48ac0c2ec595ef34dd2c294b0db)
- fix(buzz-agent): budget summarizer reasoning separately so it cannot starve the handoff summary (c7b663680a29a837dbd2fdde810f239f3d303025)
- Revert "fix(acp): reject unattended permission requests" (6a17d035f79ad582ca3f4f3cdc38d376f2c4087f)
- For the installer path, run CodeWhaleSetup.exe; it installs
- Extract the archive for your machine: codewhale-windows-x64.zip or
- Run install.bat (copies to %USERPROFILE%\bin)
- chore(release): extension 0.1.6
- Cross-surface notification message center. A backend-owned, bounded notification inbox projection now unifies task outcomes and blocking authorization requests across desktop, web, and mobile…
- Structured agent clarification workflow. Agents can now request structured multi-question clarifications through app-owned blocking choice cards with previous/next navigation, selection states…
- Authenticated GitHub skill imports. A compact GitHub token control is now available in the Import from GitHub header with verify/save, replace, masked status, failure, and remove states.
- New mobile terminal with text selection, copy, paste, and faster rendering
- Redesigned sidebar with clearer status colours and rows that are quicker to scan
- Choose what a sidebar workspace row shows: host, pull request, checks, and scripts
- Safer browser and network boundaries: sandboxed browser routes, trusted DNS targets, custom browser origins, and loopback provider endpoints now reject unsafe access paths.
- More resilient agent and provider runs: retained session writes, provider fallbacks, stream progress handling, and stdio failures now recover without silently ending active work.
- Stronger channel recovery: pending channel work resumes after recovery, acknowledgements are idempotent, and sustained Discord gateway bursts stay bounded.
- Every path is relative to the workspace root. Absolute paths, ~, .., and Windows drive/UNC paths are refused, and paths that resolve outside the root (including through symlinks) are dropped.
- Results and error messages are virtualized, so host paths are never exposed.
- Only the read-only, root-confined tools are registered: pack_codebase, read_repomix_output, grep_repomix_output, file_system_read_file, and file_system_read_directory.
- Add xai_agent_count to XaiModelSettings
- Round-trip CompactionPart through the Vercel AI and AG-UI adapters
- Add SnowflakeModel and SnowflakeProvider for Snowflake Cortex
- commit: f7491e2df17857ed959e58b69dd4bd6a1985045f
- source tag: release2026-08-08_11-14-22
- regions: cn global
- Updated bundled Claude CLI to version 2.1.226
- bb desktop 0.36.0
- Show your assistant what you're looking at. During a voice call, you can now open the camera and share what's in front of you.
- Speak in any language, and pick which one the assistant listens for. Speech recognition now defaults to multilingual, so you can switch between languages mid-conversation without the assistant…
- Each onboarding avatar now has its own voice. When you're picking your assistant for the first time, every avatar speaks with a distinct voice so you can hear the difference before you choose.
- App: fix the 0.48.0 crash-at-launch "could not load resource bundle: CodexBar_CodexBarCore.bundle" — the new provider-plugin runtime loaded its bundled scripts through the autogenerated…
- CLI: add provider/detail snapshot queries and progressive cached-shell rendering to the built-in serve dashboard.
- CLI: answer expired serve-cache requests from the last-good response while rebuilding in the background instead of blocking the dashboard for up to a minute.
Aug 7, 2026
- Kernel image-output previews. R and Python notebook runs now hydrate captured figures and saved plots locally in the renderer by compact run ID, sharing one output presentation across notebook…
- Offline detection indicator and Network status panel. A red offline indicator now appears in the home header and workspace sidebar when the network is unavailable, with a tooltip and click-through…
- Task API run-progress liveness. A provider-neutral run.progress contract with ordered lifecycle phases and elapsed time emits a ten-second heartbeat until the first visible provider output or…
- live: forward safety_settings from generate_content_config to the Live API (99c1aa3)
- avoid shell execution in ReadFileTool ranged reads and keep local environment file access in workspace (c680728)
- deps: exclude nltk 3.10.1, which breaks venvs living inside the working directory (v1) (f25bb90)
- Add support for Deepgram's Flux TTS API (v2/speak) integration
- fix(cerebras): expose max_completion_tokens
- Make every example a uv workspace member
- fix(desktop): isolate relay admission tests (74b913cff8512c015dc6f1a7473b253fa803f954)
- fix(desktop): externalize boot <style> to prevent Tauri CSP nonce override (dcc1231d6d6935819597bc42f7bc59fa0a47c8e5)
- fix(desktop): let imported and recovered identities finish onboarding (a5a9240241ad584c839f79af546df9bf0216ca1f)
- feat: add custom scan prompts and follow-up prompts
- fix: validate knowledge-base inputs during dry-run scans
- fix: make multiscan supervisor locks safe across container restarts
- Fix: four commands that failed silently while exiting 0 now surface the problem
- Fix: a Swift extension Foo in a different file from Foo no longer drops static and singleton call edges into the type
- Fix: node-id collision resolution is now deterministic and prefers active over archived paths
- Fix preservation of provider on LiteLLM-routed models.
- Harden brittle LLM event-bus mocks.
- Fix underreporting of Anthropic cache token usage.
- feat(middleware): add AgentStreamStage and AgentStreamContext types (38980ed4f)
- feat(storage-ts): accept storage as optional top-level Agent parameter (dcf09bb93)
- feat(model): add estimateUtilization method to Model base class (c84b618a3)
- feat: clarify free OpenHands model endpoints
- Fixed stuck authentication markers in the agent runtime, ensuring more stable and reliable connections.
- This update includes internal QA and packaging validation work.
- fix(agent): recover stale mutagen auth markers
No release notes published.
- Thank you @Horst1993 — for tightening agentbox configuration handling with better base URL normalization.
- Thank you @jgentes — for a conversations UX fix that stops the thread-goal footer text from auto-scrolling.
- Thank you @mwakidenis — for adding citation metadata via CITATION.cff.
- expand firebase storage rules coverage
- fix(tui): scrollbar visibility, findings scrolling, and report navigation
- chore: release v1.5.1
- UX: Typing clear in the command will match Summarize; thanks, @joestump!
- LSP: @piakdev fixed a case where LSP could panic
- MCP: Slow loading MCPs could block messages from being sent: we reduce default timeout
- docs: changelog and blog post for v0.8.6
- blog: Give GitHub Copilot CLI a memory of your codebase
- docs: add missing 0.8.6 changelog entries
- build(bazel): add credential driver targets
- feat(bazel): build vm driver and pull runtime from Github
- The conversation Explorer gains a Changes panel: see modified/added/deleted files for the project's repository, stage and unstage them inline, and get a clear result banner when an action only…
- Projects containing multiple repositories show a repo switcher with labels at the top; a single-repo project goes straight to its changes.
- Status letters (A/M/D…) are coloured by what actually happened to the file, and repository status updates live as files change on disk.
- Fixes source installs failing with ERR_SSL_SSL/TLS_ALERT_HANDSHAKE_FAILURE on machines whose npm registry is configured globally, such as corporate-managed devices behind a proxy.
- Both installers now point the portable runtime at the machine's existing npm configuration, so a configured mirror is discovered automatically.
- npm ci failures now report which registry was actually contacted and how to configure a mirror.
- Agent skill extraction is no longer stuck in a retry-then-dead-letter
- POST /api/v2/ome/trigger no longer masks strategy state. The status
- Agentic search on agent memory now uses the skill-shaped rerank
- Support channels (Feishu and Discord) in AgentScope
- Support Apple Container as workspace backend
- Support MCP & skill hubs/markets in agent service
- 734ad51c934484d6b3c65382560240004ffb64b5 MUL-5774 fix(daemon): isolate Git metadata for Windows Codex checkouts
- 3eff25401dba5e1674ee51a2db2698b89a5cb29e MUL-5785 fix(agent): return an empty output when a stream-json run has no final text
- 73ab498d1d4b8726c727d399c6db46958f20c593 MUL-5811: fix(squad) derive leader role from wire fields, not instructions text
- loopforge v0.1.0
- feat: add zlibrary provider
- feat(komari): add monitoring provider
- feat(web): add category filter to provider browser
- feat(chat-branch): persist empty branch nodes
- perf(chat-tabs): skip redundant re-show work in message tools, artifact pane, and trace page
- fix(claude-code): reap sdk cli subprocesses on shutdown
- CLI: add a built-in auto-refreshing web dashboard at / to codexbar serve.
- CLI: the serve web dashboard renders per-account sections for claude-swap providers, and codexbar serve --identity full opts authorized dashboard clients into real account emails on trusted…
- CLI: the serve web dashboard gains daily spend bar charts from local cost history, real provider brand icons served from an embedded /icons/ route, and a grouped vertical layout — multi-account…
No release notes published.
- Chronological message ordering now stays correct even when imported or legacy message IDs are out of order.
- Revert and fork actions now use real message chronology instead of message ID ordering.
- Truncation cleanup now removes stale files by file timestamp more reliably.
- Qwen Token Plan Individual — Use the built-in provider for models documented for Individual subscriptions. See API Keys.
- Authentication readiness checks — Use pi auth check to verify provider or model credentials, optionally emitting the resolved credential.
- Improved fullscreen interaction — Select words and paragraphs with multiple clicks and configure half-page transcript scrolling. See TUI Fullscreen Viewport.
- Updated bundled Claude CLI to version 2.1.224
- Windows 运行时生命周期修复:修复了在 Windows 上的 Hermes 视图守护进程的启动和状态探测问题。
- Support hiding function tools until revealed — via tool search, load_capability, or ToolReturn.tools — using each provider's native deferral/addition channel
- Add first-party run cancellation: AgentRun.cancel(), RunContext.cancel(), RunCancelled
- Add Model.resolve_prompt_cache_retention() to resolve effective prompt-cache retention from model settings
- feat(viewer): collapse secondary session sections
- fix: remove timeout fields and more
- fix(cli): reject unexpected positional arguments
- Added a Vim editing mode, off by default and switched on from /settings under Editor Mode, bringing modal editing to the prompt with Normal, Insert, Visual and Visual Line modes, a mode badge in…
- Added Vim-aware submission so a prompt can be sent without leaving modal editing, through an Editor Mode > Insert First toggle that opens the prompt in Insert mode where a bare Enter submits and a…
- Added Vim editing to the comment editors in the diff and artifact-detail views, whose footer hints follow the active mode rather than showing one fixed set of bindings, and made a collapsed block…
- Channel status reports transport and provider legs separately (@copilotkit/runtime) — status().overall === "online" proved only that the runtime reached the Gateway with a valid project key, so a…
- The in-memory agent runner is bounded — no more OOM — InMemoryAgentRunner (the default runner when the runtime has no Intelligence backend) kept every thread, every run, and a per-run…
- AGUISendStateDelta initializes arrays before /- append — a state delta appending to an array that no prior state event had initialized produced a patch that failed event compaction with…
Aug 6, 2026
- Release channels/v0.8.0
- Fix: the build_merge #479 shrink guard is no longer effectively dead
- Fix: build_merge/merge_raw_extraction prune_sources now prunes correctly when given absolute paths under a non-standard layout, deriving the scan root by suffix-matching stored source paths, and…
- Fix: graphify update now removes newly-ignored files from an existing graph
- Here is the link to our Releases
- 195c0bf39049761f095d1f1dae8a4d71e663e778
- 145dbbf1c5d1b34e5fa903392c24879b2af76906
- 8544678bf35d321e81bf31bc273272defa761a98
- Multi-window coordination: switch seamlessly between concurrent sessions without interrupting your current work.
- Image captioning: non-vision models can now generate image descriptions automatically, enabled by default.
- Vision capability is automatically detected from known provider capabilities, with no manual configuration required.
- Fixes a Windows regression that could make the recorder window grow wider on every content-height update until it overflowed the screen.
- Corrects the Advanced Protection ready status so it accurately reports on-device without claiming multilingual support.
- Adds regression coverage for repeated Windows frame-style and content-size updates.
- move casbin group rules on user rename (cdd805c)
- #2646 3c75a27 Thanks @liruifengv! - Show a cache-expiry reminder when resuming a long-idle session or submitting after a long idle stretch.
- #2697 e6e4ba2 Thanks @liruifengv! - web: When a model request fails and interrupts a conversation, a persistent failure card now stays in the session with one-click resume.
- #2652 68ba740 Thanks @wbxl2000! - Add Windows support for the built-in Kimi Computer Use capability and show the underlying error when capability setup fails. Install it from /plugins on Windows x64.
- Every in-app browser tool call used to hang for 25–30 seconds and then fail with a misleading timeout error, making the whole feature unusable. Tool calls now return properly.
- On Windows the browser failed to start at all. Fixed.
- When no browser is attached, the error now tells you what to do — open the side browser panel — instead of leaving the agent guessing.
- fix(telegram-bot): avoid getUpdates conflict on published triggers @kishanprmr
- fix(salesforce): cap Find Record FIELDS(ALL) query at LIMIT 200 @kishanprmr
- feat(azure-openai): add custom api call @AdamSelene
- 企业级认证与分布式部署:新增 OIDC、LDAP 认证,QueueFS 支持 Redis 单机、Cluster 和 Sentinel;同时补充只读副本 watch 开关、HTTP keep-alive 配置和 Redis 启动恢复机制。
- 服务端组装 Recall Context:/search 新增 mode="context",可在服务端完成配额检索、token 预算、层级降级、跨轮去重和可选摘要,客户端不再需要自行拼装上下文。
- 更灵活的资源与内容处理:资源导入新增 args.parse_mode=no_split;资源导入和内容写入新增 processing_mode,可选择完整语义处理或仅生成向量。
- Fullscreen TUI mode — Switch between regular and fullscreen modes at runtime, with a sticky editor and footer, independently scrollable transcript, and draggable scrollbars. See UI & Display.
- Mermaid and LaTeX rendering — Render Mermaid diagrams and terminal-friendly Unicode math in interactive transcripts. See Markdown settings and TUI Markdown.
- Per-directory context overrides — Use AGENTS.override.md to replace context files for a specific directory. See Context Files.
- Encode OpenAPI server variable values
- Consolidate Dependabot dependency updates
- Suppress CodeQL false positive in internal HTTP utility
- 3c9b861de306955e7d6f7a421ffb6edfbd9ef02e MUL-5765 feat(chat): starter cards under Mika's onboarding opening
- 7a7de4fa5651c6acbabf5e13b9d84112c7c47dfb MUL-5773 fix(agent): report kimi token usage from its session wire log
- b24634e6eb1d3398c9056153ffddd6aead60abba MUL-5810: fix(agent): reconcile ACP terminal usage
- @mastra/platform-workspace@1.0.0: removed MASTRA_PLATFORM_SECRET_KEY auth for workspace providers; use platform-injected MASTRA_PLATFORM_ACCESS_TOKEN or pass accessToken explicitly.
- Added item-level scorer selection for dataset experiments.
- Declarative, persistable workflow graphs.
- Custom MCP servers. Register a hosted MCP server from the WebUI and use
- IronHub install flow. Install extensions from an IronHub deep link,
- Attachments. Durable cross-channel file flows: a file sent on one
- Automatic retries are enabled by default — Non-streaming Kotlin SDK clients now apply the SDK retry policy automatically.
- Resilient Kotlin transport — Added configurable retry policies with bounded exponential backoff, jitter modes, Retry-After support, per-attempt timeouts, overall deadlines, and retry callbacks.
- Rolling pool warmups — Sandbox pool replenishment now treats warmupConcurrency as a rolling in-flight limit and immediately reuses capacity when each warmup finishes, so one slow creation no…
- ROB-770 Add cross-cluster tools documentation
- ROB-808 docs: add Bitbucket instructions for loading custom skills
- ROB-836 Pin *.sha256 manifests to LF so the Docker build works on Windows clones
- Fixed issue where Agent readiness notifications were not displaying correctly
- Fixed incorrect identity identification causing misleading notification prompts
- Added troubleshooting guide for Agent provider configuration issues
- Security Vulnerability: Fixed missing‑owner‑check at chat‑sharing link endpoint, where chat tokens could be abused to share other users’ conversation content.
- Security Vulnerability: Fixed privilege escalation allowing regular users to fetch user information of other workspaces via API.
- Security Vulnerability: Fixed privilege escalation where any user’s API Key could be used to delete arbitrary files.
- GenOffice for Linux (x64), version 0.5.149.
- bash sudo apt install ./genoffice_0.5.149_amd64.deb
- Installs dependencies automatically and adds GenOffice to the applications menu.
- 0.7.2 changelog
- [fix] Remove audio meter accessibility polling
- [feat/telemetry] Improve skill authoring workflow
- feat(viewer): add defense-in-depth security headers
- feat(benchmark): add Qwen3.8-Max results and show version inline per row
- fix(opencode): harden opencode plugin against abort, orphan, and arg edge cases
- Forward xAI FileSearchTool collections search options
- fix(azure): send max_tokens for Mistral models, not max_completion_tokens
- Omit server attributes for a malformed base_url port and forward base_url through WrapperModel
- Updated bundled Claude CLI to version 2.1.223
Aug 5, 2026
- Bump Flow canary on release
- Add URLReadTool for reading arbitrary URLs
- Add app metadata to platform action tools
- Fix: C# receiver typing no longer drops a true call when a same-named variable is declared untypeably elsewhere in the method
- Fix: graphify path (and the MCP shortest_path tool) now respect edge direction by default instead of running on an undirected view, so a returned path no longer traverses edges backwards
- Fix: semantic extraction no longer aborts at merge with a TypeError when a hyperedge carries dict-shaped members
- Simplified xAI login to a single device-code flow that works better in headless and remote environments.
- Preserved structured mid-stream provider errors so compatible providers can retry failed responses.
- Retried more transient provider and network errors instead of failing immediately.
- Godot 4.2+
- Active Ziva account
- Runs are faster and more efficient, at the same success rate — Autopilot now works over a single ongoing conversation and pulls only the page details it needs at each step, instead of re-uploading…
- Whole forms fill in one planning pass — when several fields sit on the same page, Autopilot plans the entire fill from a single look and executes the steps without stopping to reason between each…
- Clicks land more accurately — click actions are now aimed at the element's real on-screen position, reducing misclicks on overlapping or offset targets.
- Remote Control: support for CCR v2 sessions with secure worker credential exchange and real-time event streaming, synchronized with the local TUI
- Remote input integrated with the local session; slash commands and prompts processed under unified security policies
- Image preprocessing upgraded: automatic resizing and compression before upload, using the Lanczos3 scaling kernel
- License changed to MIT: The published @mem0/n8n-nodes-mem0 package is now MIT (was Apache-2.0).
- Themed icons: The node and credential icons now declare { light, dark } variants instead of a single icon, clearing the remaining icon-prefer-themed-variants warnings from the Creator Portal scan.
- Terminal: Fixed processes being left alive when the terminal was closed. (#61467; thanks feitreim)
- Fixed a hang when the Zed Agent ran a git command with a pager configured. (#61999; thanks timvermeulen)
- Fixed editor: find all references preview tabs becoming permanent tabs when any of their buffers had previously been edited. (#61021; thanks aviatesk)
- Admin config layer for shared/enforced config that applies over user config
- "Default" (unpinned) model option in /model and /config to track the recommended model without pinning a specific alias
- Server-side default model routing via experiment, never overriding explicit pins
- perf(desktop): stop gating org switch on Electric preload
- feat(desktop): add safe-triangle hover intent to sidebar hover cards
- chore(desktop): bump version to 1.18.3
- feat: set Canvas default model to GLM 5.2
- feat: activity log export
- feat: ship the featured automations landing dashboard
- test: disable tests flaky under parallel stress
- feat(build): add system CA root mode
- test(server): fix race condition in tracing tests
- AdvisorTools: Added AdvisorTools for asking advisor models for feedback.
- OpenRouteService Toolkit: Added OpenRouteService toolkit for accurate routing.
- StudioTools: Added component-aware schedule tools and history parameters.
- Review-gated session plans. A new Session Plan contract gives multi-step agent work a durable, user-reviewable execution plan.
- Hot-switch models and providers without reconnect. Changing a model or provider no longer reconnects the agent process when the target is a registered compatible option.
- Agent-aware context replay. When an ACP session cannot be resumed after switching agents or starting a replacement session, replay is now classified by target…
- Add Databricks to the list of supported databases
- feat(wren-ai-service): get knowledge from ibis (ai-env-changed)
- chore(deps): bump golang.org/x/crypto from 0.38.0 to 0.45.0 in /wren-launcher in the go_modules group across 1 directory
- a4dc54f2eda823224a4a67da8efab94fe9c2e326 MUL-5570 / MUL-5493: feat(chat): add a managed follow-up queue
- b98f760f1fab888565e4003a1ebbc4ec206ab4e5 MUL-5604: 支持 Reasonix runtime
- 79c5832e1dbcae08e8e75f6c6ca8e582629b54db MUL-5708 fix(taskfailure): classify response-side context-window overflow
- #2407 0abcd00 Thanks @wbxl2000! - Add Kimi Computer Use and Kimi WebBridge as built-in official marketplace entries in the v2 CLI.
- #2627 f881cdd Thanks @sailist! - Run the CLI surfaces (interactive TUI, kimi -p, kimi acp, kimi export, kimi provider) on the agent-core-v2 engine by default.
- #2565 54c04bf Thanks @7Sageer! - /fork no longer switches to the forked session: the current session stays active and its background tasks keep running. Find the fork in /sessions.
- macOS (Apple Silicon, signed + notarized): GenOffice-0.5.83-arm64.dmg
- Windows (x64, NSIS installer, build v0.5.79): GenOfficeSetup-v0.5.79.exe
- Web pages now open as tabs in the preview panel, and you share one persistent session with the agent — sign in by hand once, and the agent carries that session into whatever it does next.
- The agent reaches the browser through a new built-in MCP server (aionui-browser), so it can navigate, click, and pull data on its own.
- On safety: the obvious way to let an agent drive Chromium is --remote-debugging-port, but that switch is application-wide with no per-target ACL — it would expose every WebContents, including the…
- Upgraded interactions and visual design make chats, roles, projects, and tasks clearer and easier to organize
- Built for multitasking, with multi-window and split-view layouts for switching between work and reviewing progress and deliverables
- Supports both OpenAI and Anthropic endpoints, reuses Claude and Codex subscriptions, and works with more models
- GET /health now carries a cascade readiness block — healthy, human-readable reasons, and the counters behind them…
- everos cascade rebuild CLI command — drops every business LanceDB table, clears the cascade queue, and re-indexes all markdown from scratch.
- Startup schema verification now detects column *type* drift, not just missing / extra columns.
- Tool schemas now adapt to each provider automatically and recover from newly reported schema incompatibilities
- Managed cloud activation is now the default remote setup flow
- The TypeScript SDK now bundles platform-specific jcode runtimes for easier installation
- Capture post-run page evidence from the run's own browser session (SKY-14254) (b50f02566)
- Add the selector engines Skyvern's producers emit to skycdp (SKY-14071) (6a6b6383e)
- chore(deps): bump mcp to 1.28.1 to fix 3 high-severity Dependabot alerts
- 前台资源管理修复:修复了前台资源饥饿问题,并支持 Retry-After 头部。
- 召回相关性改进:改善了召回相关性和主机输入处理。
- fix: preserve completed tool guardrail results
- fix: redact invalid tool argument errors
- fix: defer non-stream session saves until output guardrails
- claude: support Claude Code in VS Code (13a310a)
- code: add PHP support to CodeAwareCompressor (6d5516d)
- compress: accept config.frozen_message_count on /v1/compress (2797099)
- Fix Google provider dropping an explicit timeout=0
- Filter unavailable names from Google allowed_function_names
- Preserve caller-owned OpenAI model settings
- feat: paste composer text without formatting (25a9cf1be6d245fbd7373cb1160dbc790baf5bd5)
- Revert "chore(release): release Buzz Desktop version 0.5.5" (79c52166cfe6b6d36bdc7686f943595c74e2f578)
- chore(release): release Buzz Desktop version 0.5.5 (a0ed13de14ee64dd90c32335790f7d3b4e94330d)
- #585 1e6b395 Thanks @colifran! - fix: stop the internal link validator from falsely flagging valid links
- #589 a86d0ba Thanks @colifran! - fix: fingerprint innermost cause and chain-walk origin tag
Aug 4, 2026
- .NET and Python: Extract Durable Task and Azure Functions integrations
- .NET: Fix Handoff orchestration sample not responding to user input
- .NET: Fail declarative workflows when an agent returns an error
- test(podman): gate gateway discovery test on Linux
- feat(bazel): add rustfmt checks to tests
- feat(vm): export driver traces over OTLP
- Updated bundled Claude CLI to version 2.1.222
- AUR publishing is temporarily delayed: the upstream AUR Git service is rejecting pushes during the active malicious-packages response. The AUR packages therefore remain at 1.21.0 for now.
- Late lock renewal failures are ignored once a run has settled — Intelligence releases the thread lock after it accepts a terminal run event, so a lock renewal already in flight could come back 409…
- @ag-ui/* unchanged from 1.66.1 — @ag-ui/langgraph at 0.0.42, @ag-ui/a2ui-middleware at 0.0.10, @ag-ui/mcp-apps-middleware at 0.0.3, @ag-ui/mcp-middleware at 0.0.1, @ag-ui/core / @ag-ui/client /…
- @copilotkit/license-verifier stays at ~0.5.0.
- Voice gets better across devices. Voice conversations now support richer controls and clearer status across web and iOS.
- Preview your files in chat: Images, PDFs, spreadsheets, and text files can now appear as previews or file cards right in chat.
- A faster model choice: Speed is now its own managed profile for quick replies, while Balanced uses GPT-5.6 Luna where available.
- Image generation: chat can now generate and edit images right in the conversation. After signing in with Grok or ChatGPT, built-in image generation works out of the box.
- Inline @-mentions now render as pills in the composer.
- Added FennoAI and Qiniu Cloud AI provider presets.
- Package contact: author.email in the published package is now integrations@mem0.ai (was founders@mem0.ai), so npm and n8n Creator Portal correspondence reaches the integrations team directly.
- New example: meeting notes → knowledge graph with SurrealDB
- fix(core): clear inverted target-owner index on component deletion
- fix(sqlite): preserve schema consistency on failed column migrations
- fix: increment previous_content with content streamed
- fix: add default fallback to instantiation.py and removed default ollama base url from components
- fix(tracing): detach orphan spans before database flush
- bb desktop 0.35.1
- Add automatic pagination for finance API responses so multi-page results (prices, financials, etc.) return complete data
- Increase finance sub-tool timeout to 60s to accommodate paginated fetches
- Paste a Jira URL to pull in an issue — kane-cli ingest <jira-url> fetches the issue and its attachments (including images) directly from Jira as a remote source, no local file download required.
- Project access is enforced before fetching — only issues in projects your account is permitted to use are accepted; unauthorized requests are rejected up front with a clear message.
- Click-then-switch-tab sequences no longer race — when a click opens a new tab, replay now waits for the tab-open popup to resolve before executing the switch_tab step, preventing the tab from…
- Branch a conversation into a new session. A split Send control now offers a Branch in new session action for any conversational session, immediately creating a local pending session that copies…
- Search GitHub skills by keyword. The skill import flow now accepts free-text keywords alongside explicit repository references.
- Specialist package import/export. Specialist profiles can now be exported and imported as versioned packages through secure ZIP and directory flows, with archive limits, validation diagnostics…
- feat(viewer): add repository search filter
- feat(allowlist): add Nix support
- feat(allowlist): add Haskell support
- GitHub pull request reviews now include the pull request number and URL in context.
- Fixed several right-to-left layout issues across tabs, drawers, resizing, and titlebar interactions.
- Fixed shared right-to-left UI behavior such as directional icons and drawer placement.
- macOS (Apple Silicon, signed + notarized): GenOffice-0.5.1-arm64.dmg
- Windows (x64, NSIS installer): GenOfficeSetup-v0.5.1.exe
- e422b169a8f6f39bd2820cd7e8b1f5795e904ee7 MUL-5415: fix(issues): keep running-agent transcript open
- 738217c2756efb18c67a5ba1989298120d60a4f2 MUL-5516: fix(issues): re-trigger blocked issues when resumed
- 55dc302d0ee34c7f580f54f75a5ddb7c814f9cb7 MUL-5610: add Codex first-item wait telemetry
- Godot 4.2+
- Active Ziva account
- npm: bash npm install -g base44@latest
- Homebrew: bash brew install base44/tap/base44
- Version: 0.1.8 NPM Tag: latest
- fix: resolve agent tool name collisions consistently
- fix: reconcile resumed tool name collisions
- fix: persist the max-turns handler output to the session
- #2558 75395f6 Thanks @sailist! - Add the TurnStarted, UserPromptQueued, TaskStarted, and SessionHeartbeat hook events, enrich hook payloads with the session title and client type, include the…
- #2416 eaab2b6 Thanks @mangeshraut712! - Fall back to the built-in models.dev catalog snapshot when the public catalog is unreachable, so Known third-party provider import still works offline or in…
- #2083 bfa0080 Thanks @StaR4y! - web: Fix dark-mode monochrome controls and align the chat composer corner radius with the design system.
- add configurable record retention days per organization (8799cfe)
- Add Bedrock support for ModelSettings.extra_headers
- Restore stable gemini-3-pro-image and gemini-3.1-flash-image Gateway aliases
- Add cost to RunUsage and cost_limit to UsageLimits
- #579 1e818ae Thanks @bracesproul! - Improve coding-agent wiki prompts and make OpenWiki guidance optional and just-in-time.
- #555 ad9c7b5 Thanks @GautamSharma99! - fix: report rejected and timed-out telemetry sends accurately
- #547 0aa6ddc Thanks @GautamSharma99! - fix: preserve agent instructions when managed markers are malformed
- cli: update gcloud command to use beta flag (9630559)
- npm plugin updates: accept singleton-array metadata from newer npm clients so tracked official plugins can install and update to correction releases.
Aug 3, 2026
No release notes published.
- Anthropic usage now shows model-specific weekly limits alongside account-wide windows
- Gemini tool schemas are sanitized for provider compatibility
- MCP notifications no longer cause request-handling failures
- ⏪ Reverts the v0.8.0 "Chat/Terminal switcher in the header" change — the switcher returns to its previous location.
- 🧹 Aligns the Omnigent desktop app to version 0.8.0.
- Currency in assistant messages stays prose — the assistant message renderer's inline-math pass matched on a bare /\$([^$]+)\$/, so a message like United at $349 … and Delta at $289 had everything…
- CopilotKit workspace packages move 1.63.2 → 1.66.0 — the published 0.3.1 pins @copilotkit/core, @copilotkit/shared, @copilotkit/a2ui-renderer, and @copilotkit/web-components at exactly 1.66.0.
- docs: add Docker/sandbox tip to installation guide
- docs: sync CONTEXT.md with new backgroundJobs and council config options
- docs(companion): add Wayland click-through known limitation
- parallel_tool_calls support: DSPy preserves each call/result pair by ID. You can do this in native mode or in non-native mode
- Multi-turn native tool call support: Prior tool calls and results can be replayed as assistant and tool messages instead of being flattened into prompt text.
- Each turn lives in dspy.History as structured messages rather than one ever-growing trajectory string, so providers with prompt caching can reuse stable prefixes more effectively.
- feat(core): add user transcription timeout
- feat(voice): auto-disable realtime server-side turn detection
- feat(cli): deprecate console and dev modes in favor of lk agent
- Added herdr --skill to print the agent skill bundled with the running Herdr binary.
- Added ui.pane_scrollbars = false to hide terminal pane scrollbars and reclaim their reserved column.
- Added ui.tab_bar_position = "bottom" to place the desktop tab row below terminal panes.
- feat: add a domain-neutral extension-manifest host
- feat(backends): compact Cloud vs Agent-server add-backend chooser
- feat: show live agent activity in chat
- #2356 04c8ee4 Thanks @Kylejeong2! - Replace raw daemon socket connection errors with a human-readable message that includes the exact browse open command needed to start the session.
- #2361 2a6e20b Thanks @shrey150! - Warn when auto-loading variables from .env and add BROWSE_LOAD_DOTENV to opt out ahead of a future release where this will be off by default.
- fix: report agent usage per provider round, not once per turn (09c86c56e52651c017743268fc8ce708bb83b265)
- fix(desktop): harden Windows installs against Defender block and orphaned Node (80315ac1a68024c40b61f3a062c9cb6bf7d4efb5)
- feat(desktop): improve channel template discovery (c1b88af8d71d1cf6aaca517e92ce9e918cd0e8bd)
- increase Provider.Scopes column size to varchar(200) (fffa920)
- Talk to Hermes — streaming, conversational voice with barge-in — Voice mode used to mean: speak, wait for the whole reply to generate, then listen to one long audio file.
- Wake words and hands-free control — Say your own open-vocabulary wake phrase ("hey Hermes", or anything you pick) and Hermes starts listening — detection runs on-device, so no audio leaves your…
- Voice on every platform — Send a voice note to Hermes on WhatsApp, Feishu, DingTalk, LINE, QQ, Photon, or Weixin and it's transcribed and answered.
- Notion AI: add Business and Enterprise workspace allowance tracking for rolling and billing-period windows. Thanks @n0ah37!
- Notion AI: pace estimates on both the rolling and billing-period bars, scored against the real calendar month ending at the reset rather than a flat 30 days. Thanks @n0ah37!
- Sync: opt-in iCloud sync (Settings → iCloud Sync, default off) syncs provider configuration, a curated preferences subset, and per-device usage snapshots across Macs via CloudKit.
- Added Business sign-in for Gemini Enterprise accounts, so you can authenticate with a Google Cloud project under Google Cloud terms, use a license seat allocated or auto-assigned from your…
- Added Workforce Identity Federation sign-in for enterprise users, available as the Use advanced SSO config option on the Google Cloud sign-in screen, so organizations that federate identity…
- Added sign-in with Application Default Credentials so you can use Agent Platform.
- refactor(otel): share OTLP trace provider setup
- feat(compute): negotiate gateway callback listeners
- fix(server): isolate otel tracing test exporters
- Project-scoped command palette. A Cmd/Ctrl+K palette now searches across sessions, files, settings, and navigation targets within the active project, so jumping to a session or opening a file no…
- Code syntax highlighting. Source files and notebook code cells now render with GitHub-style syntax highlighting through Shiki, with light/dark theme support, line numbers, and a consistent…
- Installed-package inventories. Each runtime environment card in Settings gets a Packages button with a lazily fetched count badge and a dialog showing the full name/version/build/channel table…
- 90581b083a4ea8ed09e16d3c1ff0ae949cddf0ad Merge pull request #609 from dlronray/codex/pinchtab-backlog-20260713
- d2cecbfedaa95f11525687c32737b7afe1782856 Merge pull request #611 from pinchtab/dependabot/npm_and_yarn/npm/brace-expansion-1.1.16
- ca73fadd4730305bb00864ad478d5001d7c1eb3f Merge pull request #612 from pinchtab/dependabot/npm_and_yarn/npm/js-yaml-4.3.0
- Google's Antigravity (agy CLI) is now a fully supported built-in agent. Previously it could be selected but did nothing — no chat surface appeared and the first message was never delivered.
- Long conversations get a message anchor rail along the left edge of the chat area: one tick per user message, hover to preview that turn, click to jump straight to it.
- Right-click any file in the Explorer to reveal it in your system file manager (desktop only).
- d1dd2d9 Thanks @omeraplak! - Upgrade the official Model Context Protocol SDK to 1.30.0 and add stateless Streamable HTTP handling with startHTTP, serverless, and request-scoped streaming options.
- ⏸️ Complex tasks stop halfway and require repeated prompts to continue
- 🧪 Code appears complete, but tests, builds, and runtime behavior are not actually verified
- 🔁 Regression checks, test repairs, documentation updates, and repository maintenance must be initiated manually each time
- /retry command to retry an interrupted response
- /rewind now lets you choose between forking and rewinding in place
- LLM retries surfaced in the UI and logs
- Chat Memory — 从对话中逐层提取 L0 原始记录 → L1 事实 → L2 场景 → L3
- Skill — 从跑通的任务里提炼可复用 SOP,附版本 / 资源文件 / 触发边界 /
- Wiki — 把文档变成结构化页面 + 链接图谱(灵感来自 Karpathy 的 LLM 知识库
- /caveman-stats reports net savings, not just gross. The skill costs ~1–1.5k input tokens per turn.
- Compress no longer loses data on Windows. Explicit UTF-8 plus atomic writes. A crash mid-write can no longer truncate the file it was rewriting.
- Installer moved bin/ → cli/, with hardened Windows quoting and a clean uninstall path.
- 37f3bb7dd9c0fe665051ce26dadab03b090dc1af MUL-5587: fix(autopilots): tell the user which required field blocks Create
- eb14e73a20a67929c397a644c01d6f2b47c0ceba MUL-5617: fix(storage): disable request checksum in buffered S3 Upload
- d38da27ed4cb2f9f205a00196f127ace9cd7a5e7 MUL-5619: fix(cli): surface the server's 409 message instead of the generic conflict template
- from 可重复指定多个来源目录;--to 是产物目标目录;--skill 指向 Skill 目录或 SKILL.md。
- 不加 --wait 时,命令立即返回 Task ID;也可通过 GET /bot/v1/compile/{task_id} 查询任务状态。
- 完成结果会区分 created、updated 和 unchanged,并返回页面数、链接数、警告和 OKF 版本,便于审计编译产物。
- feat(cli): add 'ocr session comments' to display saved review comments
- feat(allowlist): exclude snapshots, testdata, fixtures, and generated code from review
- fix(llm): retry once when a chat completion is truncated
- chore(gitignore): ignore local agents override file
- feat(models): support LongCat 2.0 thinking
- hotfix(ai-diagnosis): correct misclassifications and surface provider error JSON
- [fix] Avoid timeline symbol snapshot crash
- [fix] Prevent scrub reader cancellation race
- Add v0.7.0 to appcast
- fix(cloudflare): use Worker-safe validation for MCP
- fix(mcp): make provider clients Worker-safe
- fix: reduce executable registry size
No releases match.
Releases by week
Every stable release of the week, Monday to Sunday, for the last 26 weeks. Each agent's full history is on its own page.
- Nov 16, 2026431 releases · 120 agents · this week
- Nov 9, 2026717 releases · 152 agents
- Nov 2, 2026632 releases · 157 agents
- Oct 26, 2026628 releases · 145 agents
- Oct 19, 2026614 releases · 144 agents
- Oct 12, 2026545 releases · 150 agents
- Oct 5, 2026495 releases · 127 agents
- Sep 28, 2026429 releases · 121 agents
- Sep 21, 2026426 releases · 126 agents
- Sep 14, 2026299 releases · 107 agents
- Sep 7, 2026245 releases · 101 agents
- Aug 31, 2026268 releases · 99 agents
- Aug 24, 2026286 releases · 108 agents
- Aug 17, 2026234 releases · 95 agents
- Aug 10, 2026250 releases · 95 agents
- Aug 3, 2026230 releases · 89 agents
- Jul 27, 2026201 releases · 79 agents
- Jul 20, 2026199 releases · 82 agents
- Jul 13, 2026222 releases · 83 agents
- Jul 6, 2026198 releases · 81 agents
- Jun 29, 2026137 releases · 71 agents
- Jun 22, 2026157 releases · 72 agents
- Jun 15, 2026154 releases · 81 agents
- Jun 8, 2026158 releases · 72 agents
- Jun 1, 2026130 releases · 59 agents
- May 25, 2026123 releases · 67 agents