Skip to content
Agents tracked: 492 Releases (7d): 545 Agent status: 1 with issuesSince last limit reset: Claude 9d · Codex 7d For AI agents: Markdown Updated Nov 20, 2026

AI agent releases: Jun 1 – Jun 7, 2026

130 stable releases from 59 tracked agents and frameworks in the week of Jun 1, 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.

Jun 7, 2026

  • The separate light/dark toggle, color scheme, and CSS-skin systems are now a single Theme concept.
  • The Appearance settings panel (renamed from "Display") now has three independent size controls — chat text, Markdown text, and code — that apply live and persist across restarts.
  • Markdown/HTML preview now defaults to 100% zoom, matching the Cmd/Ctrl+0 reset.
  • #1613 124a47a Thanks @threepointone! - Introduce the first Think framework layer for convention-driven agent apps.
  • #1613 124a47a Thanks @threepointone! - Compile skill scripts ahead of time and remove the in-Worker bundler (drops ~14MB of esbuild-wasm from Worker bundles).
  • The Agents Vite plugin compiles bundled skill scripts (scripts/*.ts/.tsx/.js/.mjs) with esbuild at build time — resolving sibling imports and stripping TypeScript — and marks them precompiled.
Base44
  • npm: bash npm install -g base44@latest
  • Homebrew: bash brew install base44/tap/base44
  • Version: 0.0.54 NPM Tag: latest
  • AGENTMEMORY_AGENT_SCOPE=isolated not enforced on mem::search / POST /agentmemory/search / memory_recall / recall_context.
  • /graph/query and /graph/stats timed out with "Invocation stopped" on large existing corpora (#814, PR #816).
  • All data lost on agentmemory stop followed by restart. Stop order inverted: worker first with a 5s SIGTERM grace so index flush can land, engine second.
  • chore(models): prune legacy MiniMax models, keep M2.7 and M3
  • fix: encode gemini model list api key
  • fix: infer custom provider model groups
  • Full Changelog

Jun 6, 2026

  • 59edd7151e95edae2598887b22ae0f097ca873b3 feat(daemon): per-bead cargo target clean hook (hq-x0v7v)
  • 5e31df81f66b847774a41a6aef0672720d9a385c feat(witness): detect polecats stuck at startup without heartbeat (gt-uk7)
  • 64ad5ed95ea071e99a22eaa537d5d199072aee11 Revert "fix: clean standalone formula wisp on hook failure"
  • Full Changelog
MemPalace
  • Pluggable vector backends — Qdrant, pgvector, and sqlite_exact alongside the default ChromaDB, selectable via MEMPALACE_BACKEND / --backend.
  • Docker images (CPU + GPU) for the MCP server and CLI.
  • mempalace migrate-wings — one-time migration that normalizes legacy wing names (strip leading/trailing separators) so pre-upgrade palaces stay discoverable. See docs/recovery/wing-name-migration.md.
  • Markdown as source of truth: all memory persists as plain .md files you can open, edit, grep, and version with Git.
  • Lightweight three-piece storage: Markdown, SQLite, and LanceDB. No external services required.
  • Hybrid retrieval: BM25, vector, and scalar filtering in a single LanceDB query.
  • Add timeline clip mentions to chat
  • Add agent chat starter prompts
  • Render inspect_media images in agent panel and round metadata timestamps
  • Hermes Desktop — a real native app, not a terminal wrapper — This is the headline.
  • Run the desktop app against a remote Hermes — sign in with OAuth or username/password — The desktop app doesn't have to run Hermes locally.
  • The web dashboard is now a full admin panel — configure everything from the browser — The dashboard grew from "view your sessions" into a complete administration surface.

Jun 5, 2026

  • Cross-machine device pairing: vellum pair mints a device-scoped, revocable token on the assistant host and prints a pairing bundle, vellum connect import sets up a persistent paired assistant on…
  • Richer chat surfaces: new work-result, choice, and copy surfaces, plus managed-OAuth connect and approval surfaces so the assistant can hand you structured results and connection prompts inline.
  • Pro plan credit bundles: pick a credit bundle during upgrade and change flows, with the current bundle shown on the billing plan card.
  • #2192 e7d3b55 Thanks @shrey150! - Lead-with-local onboarding: the missing-API-key error on cloud commands now tells users that local browser automation needs no key and points them to browse open…
  • #1686 1e49880 Thanks @threepointone! - Batch and pack chat-persistence SQLite writes to reduce rows written and round-trips.
  • agents: ResumableStream now packs each buffered group of stream chunks into a single SQLite row (a JSON array of chunk bodies) instead of writing one row per chunk.
  • @cloudflare/ai-chat: message cleanup (stale-row pruning and maxPersistedMessages enforcement) previously issued one DELETE per row in a loop; it now deletes rows in batched DELETE ...
  • 90b1a2a67bf70a2f5fc1f998ccd84317cab5d68b: feat(server): make server prompts independent of client connections (@meowgorithm)
  • 363ffec9ec38bf84fd2476d28c6906965cda9822: feat: discover skills from git root in monorepos (@taigrr)
  • 6292bfa304c596f04caef653c47e53427ea40ca6: fix(cli): show all providers in crush models, not just configured ones (@meowgorithm)
  • HTML File Generation: Added HTML file generation support with example app.
  • AG-UI State Events: Added AG-UI state events support.
  • Tuning Engines Provider: Added Tuning Engines as a new model provider.
  • fix(grpc): allow credential rotation when legacy provider.type exceeds current limit
  • fix(cli): respect gateway name for mTLS lookup
  • chore(deps): bump actions/checkout from 6.0.2 to 6.0.3
  • feat(orchestrator): wire generate_presentation + require grounding
  • fix(inference): stream + SSE-aggregate Codex OAuth Responses calls
  • Add task source UI and agent launch flow
  • The managed runtime (Node and ACP resources) now ships inside the app package, so there's no separate download step or download/status UI on first launch — the app is ready to run out of the box…
  • Opening a .md file from the workspace tree now renders through Streamdown's native engine — proper syntax highlighting (Shiki), mermaid diagrams, and KaTeX math — and adapts cleanly to light and…
  • Fixed a regression where the markdown source view lost its color highlighting and appeared nearly monochrome.
  • cd2b067953fb7841debcc98342316ac152ef44af
  • c7cd6acbda709db6ddb044bf07deacd62475e03d
  • d92a4a6b19ea92f199812542923953805b8483d4
  • Agent Team is supported
  • fix(webui): add missing files in the web ui example
  • fix(readme): update the QR code for DingTalk group
openJiuwen
  • Enhanced evaluation module for evolution – Added benchmark integration to directly evaluate pre/post evolution effects.
  • Fixed team.stream returning no results in hierarchical_tools mode
  • Fixed temporary team not disbanding after task completion in pyzmq + process spawn_mode
  • DELETE /policy endpoint for removing egress rules — new DELETE /policy handler accepts a JSON array of target strings and removes matching rules case-insensitively.
  • Supervisor + cleanup hook — egress now runs under a dedicated single-worker supervisor (opensandbox-supervisor).
  • OTLP endpoint auto-discovery from node IP — when neither OTEL_EXPORTER_OTLP_METRICS_ENDPOINT nor OTEL_EXPORTER_OTLP_ENDPOINT is set, egress now resolves the node IP from the HOST_IP env var…
Mastra
  • None noted in this changelog.
  • Workspace sandbox now accepts a resolver function for per-request sandboxes.
  • Added untilIdle option to stream() and resumeStream() methods. Pass untilIdle: true (or untilIdle: { maxIdleMs: 60_000 }) to keep the stream open across background-task continuations — same…
  • 🔌 Cross-Process Data Plane – WsLink carries envelopes between hub and clients over WebSocket, with auth and delivery semantics.
  • 🎛️ Cross-Process Control Plane – Frame-based RPC for the hub's control surface (registration, channel lifecycle, governance).
  • New background_agent_tool for running an agent in the background while the parent agent continues — useful for long-running work that shouldn't block the foreground conversation.
  • docs(bot): document gateway.token (required when host is non-localhost)
  • docs(quickstart): note Web Studio /studio is available from pip/pipx server installs
  • feat(cli): improve config wizard and result UX
  • ai edit menu on right click a media
  • Show generating state on timeline clips instead of offline
  • no full screen for home page

Jun 4, 2026

  • agent-framework-core: Add MCP-based skills discovery (McpSkillsSource)
  • agent-framework-core: Progressive tool exposure via FunctionInvocationContext
  • agent-framework-core: Add background agent support to harness agent
  • agents: LlmAgent default model changed from gemini-2.5-flash to gemini-3-flash-preview (a preview model), ahead of the 2026-10-16 gemini-2.5-flash shutdown.
  • interactions: GenAI SDK v2.0.0 support renames the turn-based helpers in interactions_utils.py (e.g.
  • Add --trigger_sources and ADK service options to cli_deploy_agent_engine (ffa057c)
KaneAI
  • New kane-cli generate "<objective>" command — starts an AI-driven generation session in an interactive TUI.
  • Headless / scripted generation — run kane generate with --refine or --save flags for non-interactive pipelines; --out controls where the resulting files land.
  • Scenarios drill-in with /view — while generating, /view opens a full-screen browser showing scenarios and individual test cases as they are produced; cases can be excluded before saving.
  • chore(mise): refresh tool lockfile
  • ci(release): authenticate snap canary artifact download
  • docs(container-gateway): fix Docker driver setup for containerized gateway
  • More built-in provider coverage - Added Ant Ling and NVIDIA NIM provider setup, plus MiniMax-M3 support for the direct MiniMax providers. See Providers.
  • Richer extension context - Extensions can use ctx.mode and ctx.getSystemPromptOptions() to adapt behavior across TUI, RPC, JSON, and print modes and inspect base system prompt inputs. See Extensions.
  • Added containerization documentation and a Gondolin extension example for routing built-in tools into a local micro-VM.
  • Document query index reuse — Repeated document_query calls within the same chat context now reuse an existing vector DB instead of re-parsing and re-embedding the document every time, while…
  • Folder attachments in chat composer — Dropped or selected folders are flattened into ordinary file attachments, preserving basename display and existing multi-file upload behavior.
  • Browser tab reuse guidance — The Browser tool prompt and plugin docs now instruct agents to list and activate existing tabs by URL or title before opening new host-browser tabs.
  • Image attachments via @-mentions in the TUI for vision-capable models
  • Session deletion, exposed over ACP as the session/delete extension method
  • Browser sign-in now shows a copyable fallback URL when the browser does not open, so SSH and remote sessions can complete onboarding
  • ci: add managed node mirror workflow
  • feat(workspace): VSCode-style file tree icons + smoother preview browsing
  • feat: add managed acp artifact mirror workflow
  • copilot: GitHub Copilot subscription mode through Headroom (f4dff9b)
  • ccr: scope proactive expansion by workspace (cross-project leak) (197601b)
  • ccr: scope proactive expansion by workspace (cross-project leak) (1bc163f)
  • None noted in this changelog.
  • Add a type-safe MessageFactory component to @mastra/react for rendering a MastraDBMessage with your own per-part components, and align its tripwire/task-verdict types with @mastra/core.
  • The tripwire reason is now persisted as tripwire.reason (previously it was only stored in the message text part).
  • feat(ai): add Mistral AI as a platform AI provider @onyedikachi-david
  • feat: formula & data manipulation functions in flow builder @Louai-Zokerburg
  • feat: Platform admins can now view run success rates, live queue depth, and 30-day health history in one dashboard @MrChaker
  • wren: query MDL views in the SDK rewriter and strict mode (731c9c0)
  • wren: serve agent skills and reference docs from the CLI (cbd10cd)
  • wren: support Databricks catalog (5e58ef6)
  • fix(SKY-10653): stop browser profile dropdown clipping its last item (d74864843)
  • feat(SKY-9957): clarify Browsers vs Browser Profiles subheader copy (2b8db6a23)
  • SKY-10578: Add credential folders (frontend) (817cc9f07)
  • Custom sandbox network CIDR: Users can now specify CUBE_SANDBOX_NETWORK_CIDR at install time to avoid conflicts with existing host network subnets.
  • Systemd 255 ExecStart/ExecStop compatibility: Prefixed all ExecStart, ExecStartPre, ExecStartPost, and ExecStop directives with /usr/bin/bash to avoid 203/EXEC errors from systemd-executor on…
  • glibc version preflight check: The installer now checks glibc >= 2.34 before installation to fail fast on unsupported distributions (e.g., CentOS 7, Ubuntu 20.04), where cubelet/cubecli binaries…
Crawl4AI
  • PyPI: bash pip install crawl4ai==0.8.9
  • Docker: bash docker pull unclecode/crawl4ai:0.8.9 docker pull unclecode/crawl4ai:latest
  • Note: Docker images are being built and will be available shortly. Check the Docker Release workflow for build status.
  • Homepage: Added workspace homepage statistics, enabling administrators to overview all workspace resources, operation monitoring trends and resource usage rankings.
  • Knowledge Base: Added customizable terminology library; custom terms are prioritized during document word segmentation and retrieval.
  • Agent: Full multimodal input support for AI Conversation nodes.
  • bb desktop 0.0.15
  • For the installer path, run CodeWhaleSetup.exe; it installs both binaries under %LOCALAPPDATA%\Programs\CodeWhale\bin and adds that directory to the current-user PATH.
  • Extract codewhale-windows-x64.zip
  • Run install.bat (copies to %USERPROFILE%\bin)

Jun 3, 2026

  • Python: Improve function call invocation parameter consistency
  • Python: [Breaking] Update OpenAPI document parsing options
  • ci: harden Python test coverage workflow
Herdr
  • Fixed a Herdr server crash triggered by pane output containing complex Unicode, emoji, or decomposed accent graphemes.
  • Direct installs managed by mise now update through the mise install path instead of failing to replace the active binary.
  • Claude Code panes that are actively thinking or streaming no longer flicker to blocked because of custom status text.
  • afa7834e2ec8a93b2224fe7ab184b97fbcaa8c9a Updating dotnet package versions for 1.9 release
  • c6951c21f69af4c9a5d5cbe107b36fca9f648907 Python: Add MCP-based skills discovery (McpSkillsSource)
  • a9824289168c5af4acc5bfc7bd227206eb942fdc .NET: Bug fixes for AGUI hosting and workflows
Stagehand
  • Updated dependencies \[3a53ed4, 6e75725, 8fc16d2, 78bcde8, 3e95a87, ebbdcd3, 12703a6, 1db5f1c, cb586a1, 765861c, 2cd60a3, e102a89, 49575d6, dc1445d]
  • @browserbasehq/stagehand@3.5.0
  • Add spawn_subagent to delegate focused sub-tasks to isolated workers (general-purpose, research, analysis); emit multiple calls in one turn to run independent tasks in parallel
  • Add CLI subagent progress UI with per-delegate status, tool-use counts, and rolled-up activity
  • Improve web_fetch with Readability/Turndown HTML-to-markdown, HTTP caching, and fast-model extraction from fetched pages
  • None called out in this changelog.
  • Added channels.threadContext.addSystemMessage to opt out of the built-in channel system message.
  • Added agent.sendMessage() and agent.queueMessage() APIs for sending user-authored input into agent threads.
  • xAI SuperGrok OAuth subscription provider
  • Replay ACP images on session load
  • Expose raw provider supported models over ACP
  • (web) plumb temperature through Responses API
  • (engine) scope v1 history for channel conversations
  • (release) add WeCom release artifact
  • 桌面底层切换到 Electron:macOS / Windows / Linux 三端的内置浏览器、Workbench 预览、终端、通知、托盘、菜单和窗口状态全部迁移到新桌面运行时,跨平台行为更统一。
  • 内置浏览器与 Workbench 适配新壳:预览、截图、网页元素选择和本地文件打开链路在 Electron 壳下保持完整并更稳定。
  • Agent 指令体验打磨:agent slash 走正常聊天链路并流式展示进度,选择气泡更稳定,会话标题按用户语言生成。
  • Full Changelog
  • feat(providers): add Google Vertex AI inference provider
  • fix: correct example paths in local-inference README
  • ci(release): bring Fedora RPM canary to parity
  • #1659 f99f890 Thanks @threepointone! - Recover one-shot scheduled work (alarms) killed by a "This script has been upgraded…" deploy/code-update, not just "Durable Object reset because its code was…
  • #1661 41315b6 Thanks @threepointone! - Enforce the tool_use.input invariant at the chat write boundary.
  • #1665 13d6db0 Thanks @threepointone! - Await Chat SDK state-agent cleanup scheduling during startup so tests and short-lived worker isolates do not leave dangling cleanup work.
  • refactor: improve matrix parsing validation and performance
  • fix(inference/ollama): use prompt-guided tool calling so skills work on Ollama (#3098 sub-issue 3)
  • feat(memory-tree): per-integration health strip
CodeWhale
  • For the installer path, run CodeWhaleSetup.exe; it installs both binaries under %LOCALAPPDATA%\Programs\CodeWhale\bin and adds that directory to the current-user PATH.
  • Extract codewhale-windows-x64.zip
  • Run install.bat (copies to %USERPROFILE%\bin)
  • DALLEImageGenerator has been updated to account for OpenAI's retirement of the DALL-E models. The default model is now gpt-image-2 (previously dall-e-3). To migrate
  • Update model value: besides gpt-image-2, gpt-image-1 and gpt-image-1-mini are also supported.
  • Update quality value: the new accepted values are auto, high, medium, or low (previously standard or hd).
  • Warp Stable release
  • First boot crash: TypeError: Cannot read properties of undefined (reading 'v'). The sharded index load path checked manifest.value !== null before forwarding, but some iii-state adapters return…
  • 125 test files / 1381 tests pass
  • Build clean
  • 原生 ov CLI 体验重构:ov config 现在是配置管理入口,可交互添加、编辑、删除、切换配置;ov config show、ov config validate、ov config switch 保留为显式子命令。新增 ov language / ov lang 选择显示语言,ov status [--verbose] 提供聚合诊断视图,ov health 与错误提示改为更可读的渲染。
  • Web Studio Playground 与身份管理:Studio 侧边栏新增 Playground,可在同一工作区内查看上下文树、运行 Terminal 操作并与 Agent 面板交互;Connection & Identity 页面支持保存连接、选择 account/user/agent、创建 account/user、复制或重新生成 API key。
  • VikingBot 经验召回配置化:新增 bot.ov_server.recall_exp_first_round_only、exp_recall_limit、exp_recall_max_chars,用于单任务/评测场景中只在第一轮注入 agent experience;本地与远端模式都按传入 agent_id 做经验命名空间隔离。
  • generating ui
  • Open AI access to free tier, gated by credits
  • Anchor text clip resize to alignment edge
  • feat(gradium STT): add language option to gradium.STT
  • fix(workflows): bind GetDtmfTask user-state handler to user_state_changed
  • fix(examples): check anam_avatar_id in ANAM_AVATAR_ID guard
  • Improved reliability of web release builds, ensuring shared packages are correctly installed before the build process runs
  • Fixed a configuration issue that caused web release builds to not properly enable platform mode, resolving potential runtime errors in the web application

Jun 2, 2026

  • Cross-provider fallback always 404'd and tripped the circuit breaker. Fallback resolved each provider's own env-driven default model instead of inheriting the primary's.
  • import-jsonl aborted entire batch on legacy session row missing id. Re-keys on parsed.sessionId, backfills missing existing.id.
  • parseSummaryXml silently dropped summaries wrapped in markdown fences. New stripXmlWrappers() peels fences + pre/postamble; final-merge parse retries once.
Openhuman
  • Add vault file-system watcher implementation
  • fix(subconscious): degrade WAL journal mode to fix Intelligence tab hang on shm-incompatible filesystems
bb
  • bb desktop 0.0.14
  • f6b8592a5acaffa144dd35c2cd78f24fa270b77e: feat: prepare alibaba migration to /messages (@andreynering)
  • 3cd88ac8902c1ce421d9a2b32ae0ef7b9578d35e: fix(agent): add 401 retry and reauth notification to sub-agent runs (@taciturnaxolotl)
  • 90197756b670d743ad3c84528e4d69d2884a057a: fix(agent): centralize 401 retry logic and fix notify-on-success bug (@taciturnaxolotl)
LangChain
  • Changes since langchain==1.3.3
  • release(langchain): 1.3.4 fix(langchain): improve HITL rejection guidance
  • Assert on network requests directly in your tests — KaneAI can now inspect live network traffic during a run, letting you assert on request/response bodies, status codes, and headers using…
  • Read and assert on browser console output — console logs (errors, warnings, app-level messages) are captured per run and can be used as assertion targets or extraction sources, with level…
  • Cookies and localStorage are now inspectable — Kane-cli can read, assert on, and extract values from cookies and localStorage during a test run.
  • release(langgraph): 1.2.4
  • test(sdk-py): add factory-graph integration test exercising the server factory path
  • fix(langgraph): keep _on_started backward-compatible with overrides predating cause
  • Renamed Remote Link to Remote Control with a refactored tunnel provider system. Cloudflare Tunnel, Microsoft Dev Tunnels, Serveo, and Tailscale are available through the Remote Control selector.
  • Advertise Agent Zero version in connector handshakes so CLI clients can detect version mismatches with the connected server.
  • Expose installed plugin toggles via a protected API endpoint that lists installed plugins and supports toggling, while refusing changes to protected plugins like the connector.
  • fix(agent): Force streaming on Agent LLM so tokens stream incrementally
  • fix: change docling remote component to receive task id and to throw processing errors
  • fix(ci): lockstep langflow-nightly and langflow-base-nightly versions (forward-port
  • feat(telemetry): add anonymous opt-out OpenShell usage telemetry
  • ci(release): gate helm/oci artifact publishing on release
  • ci(kubernetes): stabilize HA e2e setup
Agno
  • Parallel Web Task API & Monitor API Tools: Added tools for Task API and Monitor API integration for parallel web.
  • Manifest: Added Manifest for per-entity AgentOS UI metadata configuration.
  • WhatsApp: Upgraded WhatsApp Graph API version to v25.0.
  • docs: add 0.7.1 changelog and release blog post
  • chore: regenerate docs-skill references after v0.7.1
  • docs(multilingual): add pg_search backend to BM25 selector and comparison table
  • Models now stay in sync after recovering a conversation — the chat header and send box show the same model, and stale handshake models are no longer restored when a session is briefly unavailable.
  • Restored conversations wait for the ACP runtime to finish warming up before loading model and mode state, eliminating the early "404" errors that could appear right after reopening a chat.
  • Switching an ACP model or mode now gives clear, localized success and failure feedback.
  • #1623 4c8b371 Thanks @threepointone! - agentTool() now returns a structured failure envelope instead of an opaque error string, so a parent agent can tell a transient interruption apart from a…
  • interrupted → retryable: true (the run never reached a logical outcome; re-dispatching can succeed), and now surfaces the underlying interruption reason via error.
  • aborted (intentional cancellation) and error (genuine failure) → retryable: false.
  • CubeCoW Copy-on-Write snapshot engine: A full-lifecycle CoW snapshot engine using reflink-based volume snapshots, providing efficient block-level snapshot operations for sandbox volume management.
  • Soft-dirty incremental memory snapshots: Per-cycle incremental memory snapshots via soft-dirty page tracking, dramatically reducing snapshot time and storage for repeated snapshot cycles.
  • Snapshot restore & vsock handling: VSOCK connections are properly reset on snapshot restore, ensuring clean connection state after rollback.
  • For the installer path, run CodeWhaleSetup.exe; it installs both binaries under %LOCALAPPDATA%\Programs\CodeWhale\bin and adds that directory to the current-user PATH.
  • Extract codewhale-windows-x64.zip
  • Run install.bat (copies to %USERPROFILE%\bin)
  • fix: guard OV OCR availability CPU detection
  • fix: respect topic naming setting for agent sessions
  • feat(models): add MiniMax M3 to the system model registry
  • .NET: Persist ForeachExecutor iteration state across checkpoints
  • .NET: Remove responses experimental flag from FoundryAgent et.al.
  • .NET: Add MCP-based skills support (skill-md type)
  • export .palmier
  • Redesign export dialog: bigger, titled panels, Format dropdown
  • Add "Export as Palmier Project" (self-contained bundle)

Jun 1, 2026

  • Fix bug where grounding metadata in Gemini 3.1 live was being silently discarded (9b6b9e9)
  • fix input and output transcription finished events for Gemini v3.1 (13763d7)
  • tools: Prevent session drop on MCP tool error (1fd406b)
  • fix(cli): wrap CLI breadth — cline, continue, goose, openhands
  • fix(subscription): wire tokens_saved_rtk data plane
  • fix(observability): RTK metrics + Rust observability (Phase H blocker)
  • refactor(proto): move phase and current_policy_version into status
  • feat(python-sdk): support OIDC Bearer auth on SandboxClient
  • fix(helm): vendor chart dependencies before release packaging
  • Improved vitals command output formatting for better readability
  • Surfaced agent-browser feature coverage in documentation
  • release(langgraph): 1.2.3
  • feat(langgraph): wire RemoteGraph.interleave to sdk-py interleave_projections
  • feat(langgraph): add v3 streaming support to RemoteGraph
  • feat(agent): multi-agent personalities with scoped memory and delegation
  • feat(agent): add orchestration control plane
  • feat(agent): live streaming subagent view with reopenable sub-threads
  • Connector-first broker profiles (IBKR + Robinhood). Trading access starts from a selectable connector profile instead of separate broker/live entry points.
  • Research Goal runtime. Long-running, research-only goals with auditable checklist criteria, budgets, and a /goal CLI command, plus REST + MCP endpoints and a Web GoalDrawer.
  • Swarm pass. Live reconcile + MCP keepalive, operator-configured external MCP tools in workers, DAG gating when an upstream task fails, a strict alpha-bench random control, and a new retry_run to…
  • PyPI: bash pip install crawl4ai==0.8.7
  • Docker: bash docker pull unclecode/crawl4ai:0.8.7 docker pull unclecode/crawl4ai:latest
  • Note: Docker images are being built and will be available shortly. Check the Docker Release workflow for build status.
  • Skills Hub can now import a skill from a folder, a parent folder, a SKILL.md file, or a zip archive — not just a single fixed format.
  • Sending a follow-up while a response is still streaming no longer shows a generic "Request failed" card with a misleading Retry button.
  • The local web backend now skips ports that the system blocks for fetch, so startup is more reliable, with added logs for the selected port and health readiness.
  • Spawn Subagent: New spawn_subagent tool for ephemeral in-workspace sub-agent execution
  • Open Directory: New "Open Directory" tab to reference local projects without copying files into the workspace
  • Tencent Yuanbao Channel: New channel integration for Tencent Yuanbao bot
  • Extract codewhale-windows-x64.zip
  • Run install.bat (copies to %USERPROFILE%\bin)
  • Add %USERPROFILE%\bin to your PATH
  • Update azuredevops_provider.py
  • fix: prevent KeyError when removing type from PR description data
  • fix: return empty list on ValueError in _get_commands_list_from_settings
  • WebUI as a daily workbench — The WebUI is no longer just a packaged chat surface.
  • Long-running agent stability — Goal mode and multi-turn work got sturdier: sustained goals keep the runner alive, goal iteration budget is extended, process_direct now has per-session locking…
  • CLI Apps, MCP, and extension workflows — CLI Apps landed as a way to connect CLI-native tools, then got unified with MCP concepts through preset setup, capability mentions, stale install recovery…
  • Telegram Bot: Connect a Telegram bot with support for text and multimedia messages
  • Discord Bot: Connect a Discord bot to chat in channels and direct messages
  • Slack Bot: Connect a Slack bot and bring CowAgent into your team's workflow
  • Three-tier progressive-disclosure navigation. The shell exposes exactly four primary surfaces that are always visible — Home, Runs, Workspace, Workflows — backed by a single nav registry in…
  • Registry-level developer-mode gating. Developer surfaces are gated by conditional construction, not CSS.
  • WELCOME → FOCUS → DETAIL altitude model. The app has three navigational altitudes you can always locate yourself in.

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.