AI agent releases: Jun 22 – Jun 28, 2026
157 stable releases from 72 tracked agents and frameworks in the week of Jun 22, 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 28, 2026
- init no longer crashes on a detected agent that has no SDK constructor mapping yet.
- smithers ps returns an empty list in a never-initialized directory instead of a PS_FAILED error. A fresh directory simply has zero runs, the same way inspect and output treat a missing store.
- smithers hermes prints a clean result instead of a nested object dump.
- 🌐 Multi-provider browsers — new cloak browser support and a provider
- 🥷 Stealthier automation — silent-drop recovery with stealth-score
- 🎮 Interactive screencast — held-modifier keyboard chords, clicks scaled
- Kimi CLI --print compatibility (#1461 fixing #1456, @chenhg5, reported
- Newer kimi-code 0.14.x removed the --print flag, causing error: unknown option '--print' (Did you mean --prompt?) on agent startup.
- cc-connect now probes Kimi CLI --help at startup and conditionally passes --print only when the installed binary supports it.
- npm: bash npm install -g base44@latest
- Homebrew: bash brew install base44/tap/base44
- Version: 0.1.0 NPM Tag: latest
- #1826 1bbd9bc Thanks @threepointone! - Add a tight, OOM-specific retry budget to chat recovery so a memory-limit crash loop seals fast and attributably.
- #1826 1bbd9bc Thanks @threepointone! - Fix neverending chat-recovery retries when a Durable Object isolate runs out of memory mid-turn.
- feat: Add admin API endpoints for runtime config reload
- Add firewall detection and user-friendly error handling for Supabase sign-in
- operator mode: add event-driven triggers
- 0.4.5 changlog
- [timeline] Add select forward and ripple insert preview
- [timeline] drag to reorder tracks via header grip handle
- Godot 4.2+
- Active Ziva account
- Full Changelog
Jun 27, 2026
- Full Changelog
- Client: Add expirationDate to AddMemoryOptions, update(), and the Memory interface; add showExpired to SearchMemoryOptions and GetAllMemoryOptions
- LLMs: Add MiniMaxLLM provider backed by the OpenAI-compatible MiniMax API (api.minimax.io/v1, default model MiniMax-M2.7)
- LLMs: Add LiteLLM provider for routing requests through a local or hosted LiteLLM proxy
- Initialize Git repositories for generated projects
- Require explicit CrewAI project definitions
- Open deployment page after CLI deploy
- 🤖 Seven new agent harnesses: Hermes (SDK + native TUI), GitHub Copilot (SDK), OpenCode, Goose, Qwen Code, Kiro, and Kimi Code; plus Antigravity promoted to a full SDK + native agy CLI harness
- 🪢 Native-harness parity across the fleet: compaction, cost/token tracking, resume, true fork with history, in-session model switching, and tool-approval / AskQuestion web cards, for cursor, codex…
- 🖥️ Omnigent Desktop: now manages the server + runner automatically
- Hermes, now built in (local & remote gateway) - Hermes is a first-class adapter in this release.
- Task watchdogs - A first-class watchdog control plane lets you attach automated checks to a task and have Paperclip watch it for you — surfacing watchdog state, outcomes, and outstanding work…
- Ask work mode - Issues can now run in an "ask" work mode for question-and-answer tasks, so you can point an agent at a question and get an answer back without it spinning up a full execution…
- [export] fix fcpxml and supports v1.10 - v.1.14
- Merge pull request #194 from barissalihbabacan/fix-turkish-readme
- Add v0.4.3 to appcast
Jun 26, 2026
- fix(desktop): split lib modules under size guard (e7d43dc2)
- Fix desktop notifications on GNOME 46+ Linux (ca50d832)
- perf(desktop): debounce channel-list refetch + profile get_channels (c6e3e947)
- title: "0.25.4" description: "A focused patch from a post-publish smoke test of 0.25.3: backend resolution honors a reverse-migration receipt, the Get Started first-run command works everywhere…
- Smithers 0.25.4 is a focused patch driven by a post-publish smoke test of 0.25.3.
- After a reverse migration (bunx smithers-orchestrator migrate --to sqlite), a default bunx smithers-orchestrator ps or inspect no longer refuses with SMITHERS_BACKEND_CONFLICT.
- feat(deepseek): register deepseek-v4-pro and deepseek-v4-flash models
- feat(models): support Databricks-hosted models
- fix(azure): decode percent-encoded spaces in Azure DevOps PR URLs
- feat: Add full Git history user setting
- feat(enterprise): add admin user-provisioning endpoint
- feat: Use interrupt endpoint for agent pause UI
- Welcome Martian! Thank you for #3941 fix(composio): pin toolkit_versions=latest in direct-mode tool listings; #3946 fix(tauri): identify foreign process on core port + consent-gated force-quit.
- Welcome Shabnam! Thank you for #3970 fix(sentry): re-register inboundFiltersIntegration so ignoreErrors runs; #3984 fix(settings): render Managed provider as always-on badge, not a locked toggle.
- Cyrus Gray — thank you!
- Fixed the React renderer so it picks the react-dom renderer instead of hardcoding renderer id 1, which prevented reading an empty tree on Next.js 16.3 Turbopack
- Full Changelog
- fix(snap): use snap-owned XDG directories
- chore(deps): bump azure/setup-helm from 5.0.0 to 5.0.1
- refactor(cli): replace sandbox_create positional args with SandboxCreateConfig struct
- AionUi's built-in Butler can configure assistants, skills, MCP servers, and models, set up remote access, and diagnose problems in plain language — but it was hard to discover.
- The bug report dialog gained a "solve via chat" button — it bundles your problem description and screenshot into a prompt and sends it straight to the Butler for troubleshooting.
- The clock icon in the conversation header no longer appears for chats that have no scheduled task, removing clutter for users who don't use cron.
- Fallback handling for Responses API rejections — OpenAI-compatible providers that reject /v1/responses payloads with a 400 validation error now correctly retry through Chat Completions.
- Materialize MCP image attachments — MCP image and image-resource payloads are now saved as scoped artifact files, returning real file paths in tool text and attachment metadata instead of only…
- Normalize Responses tool schemas — Function parameter schemas now include an explicit properties object before Responses requests are dispatched, keeping permissive prompt/MCP schemas compatible…
- ClickhouseDB for Traces: Added support for ClickHouse for high-volume trace ingest and OLAP scans. See docs.
- Scavio Search Toolkit: New search toolkit integration with Scavio.
- OpenAI: Citations from web-search responses are now surfaced on response.citations for OpenAIChat and OpenAILike providers.
- New truth subspace: builds centroids and slots from distilled session lessons to rerank and align search results with what the system 'learned'.
- Truth-subspace reranking + feedback activation: retrieval can now use a learned feedback signal to improve result relevance (opt-in by configuration).
- Improvements to integrity and filtering: signatures now use sha256 and centroid session filtering has been tightened to reduce noise in truth anchors.
- #1758 6b46b04 Thanks @threepointone! - Add progress signalling and durable milestones for agent-tool sub-agents
- reportProgress({ fraction?, message?, phase?, data? }, { persist? }) on
- onProgress(run, progress) parent hook, fired best-effort from the tail
- Windows: 下载 GenericAgent-Desktop-Windows-Portable.zip,解压后双击 GenericAgent.exe 启动。卸载时先退出应用,再双击包内 uninstall.bat,最后删除整个解压目录。
- Linux: 下载 GenericAgent-Desktop-Linux-Portable.tar.gz,解压后执行 chmod +x GenericAgent.AppImage,再运行 ./GenericAgent.AppImage。卸载时先退出应用,再运行 ./uninstall.sh,最后删除整个解压目录。
- macOS: 下载 GenericAgent-Desktop-macOS.dmg,双击打开 DMG,把 GenericAgent.app 拖入 Applications,之后在“应用程序 / Applications”中启动。若出现“无法验证开发者”等提示,先双击 DMG 内的 open_anyway.command,再启动。卸载时先退出应用,再从 Applications 删除…
- bbf758e1afd7e48dc2acdf692f7f09fe3d7e82fd docs(changelog): add v0.3.31 entry for the 2026-06-26 release (MUL-3748)
- 8c84415864064ece9d798ccd699efdb29b1c8a1f docs(claude): note pnpm dev:desktop self-isolates per worktree
- 8d0ea04fb061b2e95beb699a4cc43a016242a039 feat(composio): add standalone Go SDK client (MVP)
- 0.4.3 changelog
- Add FCPXML 1.14 export
- [ui] double click a clip on preview selects that clip
- fix(codex): adjustable reasoning levels (low/medium/high/xhigh) in picker (4745242)
- fix(tui): repair the last 4 red tests, CI now fully green (c3c796d)
- fix(tests): make CI deterministic — hermetic fixtures + serial test run (84d4840)
- 🖥️ A2UI – A2UI joins the beta framework! Beta agents can now generate UI on the fly. Docs
- 👁️ Observable run surface – New Agent.run and AgentReply.run methods sit alongside ask, returning an observable handle so callers can step through and/or watch the run as it unfolds. Docs
- 🔁 Agent Resume – Resume an agent's execution from an intermediate state, rather than just a turn. Allows fine-grained replays and forking. Docs
- A2A agent discovery: agents can describe and discover one another over the gateway, opening up agent-to-agent interop.
- Richer skills story: user-configured extra skill registries via registry:<name>/<skill> and typed slash-command options in SKILL.md frontmatter.
- Installation now adds zeroclaw to PATH automatically with a --no-modify-path opt-out.
- wren: add v5 project layout (schema_version 5) (d50b9bc)
- wren: case-aware column & model binding in the CTE rewriter (58f3f51)
- wren: v5 project layout — knowledge/ first-class, memory decoupled from LanceDB (3e34906)
Jun 25, 2026
- Unlock Chat History Cap
- fix: API update-embeddings fails with Prisma "filename is missing" on Windows paths
- Intelligent Tool Selection (ToolReranker) enabled by default
- Track conversational flow turn usage in telemetry
- Support conversational flows in the CLI TUI
- Add unified declarative flow loading
- d9ec5eaab4918612ae0f7ddf11499576bdb510bf update package version
- e3b64fdc4749256fa2d559be18a41f1a008dd7f6 .NET: [BREAKING] Make all AgentSkillsProvider tools require approval by default [ #6727 ]
- 3a5bbb5f8e3991d06fc3fe67fe3b01f2441cab0a .NET: Add DeclarativeWorkflowJsonOptions for AOT-safe declarative workflow checkpointing
- ACP method for managing recipes and UI changes
- Use ACP to manage global config and session extensions
- Recipe handling in new_session
- Toggle WebSocket capture from the TUI — a new switch in Config > Run lets you turn WS frame capture on or off without editing config files.
- WS frames appear alongside network activity — WebSocket traffic is folded into the devtools network view, so HTTP and WS events show up in one place during both test runs and authoring sessions.
- No more invented response field names — when extracting values from API responses, kane-cli now insists on named extractions tied to real fields rather than making up key names.
- For the installer path, run CodeWhaleSetup.exe; it installs codewhale.exe, codew.exe, and codewhale-tui.exe under %LOCALAPPDATA%\Programs\CodeWhale\bin and adds that directory to the current-user…
- Extract codewhale-windows-x64.zip
- Run install.bat (copies to %USERPROFILE%\bin)
- Restore workflow - Added --restore, --restore-save, restore validation flags, worktree-scoped session id, session info, and --namespace so agent runs can use stable, isolated, automatically…
- Hardened session lifecycle handling with explicit daemon and browser compatibility checks, lifecycle status output, MCP support for restore options, and safer auto-save behavior that avoids…
- Fixed restore lifecycle edge cases around switching restore keys with a live browser, daemon configuration startup races, launch mode validation, and clearing restore failures after an explicit…
- bb desktop 0.0.24
- #2246 303ab2c Thanks @shrey150! - browse screenshot now writes a file by default instead of printing base64 to stdout.
- #2250 8b83bb7 Thanks @shrey150! - Fix browse skills add on Windows and bound the unbounded installer stages.
- Quote the npx command and arguments when spawning through cmd.exe (shell: true for .cmd/.bat shims), so the default C:\Program Files\nodejs\npx.cmd path and install paths with spaces (e.g.
- feat(desktop): refresh Agents tab live on inbound relay sync (9a59e308)
- fix(buzz-acp): inject Codex network allowlist for relay hostname at spawn time (0379247e)
- refactor(desktop): consolidate notification helpers and add channel names to toasts (dd5592e3)
- 0.4.2 changelog
- fix(export): surface XML/FCPXML write failures instead of silently succeeding
- Agent: set_project_settings tool + auto-match timeline on clip add
- Guided onboarding in the web app. New users get a guided flow after signup that suggests plugins based on how they plan to use Vellum, connects Google Calendar, and schedules a Day 2 check-in.
- MCP server management. MCP settings have moved under Integrations. You can now view connected MCP servers (including which tools are registered on each and the token contribution) and remove…
- Plugin publishing. A new assistant plugins publish CLI command lets authors publish to the marketplace directly.
- c1a482267431964b9c27f16b925fd659a1b222c2: feat(providers): add llamacpp enricher (@taciturnaxolotl)
- 4be77c5675b38f851a94040794cbe3417f5dee20: feat: log provider warnings from fantasy step results (@taciturnaxolotl)
- de6792039e757b533589da4165a2ed4182ed94cb: fix(auth): stop two Crush instances from invalidating each other's login (@taciturnaxolotl)
- Full Changelog
- Full Changelog
- Run hydration — runs now hydrate from RunStore and persist interrupted
- Custom agents that update themselves — agents can persist edits to their
- User-owned IM channel connections — users can bind their own Slack,
- docs: add step for creating the GatewayClass
- test(e2e): stop using custom e2e binary builds
- fix(e2e): make postgres fixture compatible with OpenShift
- CLI-Matrix — a new cli-hub matrix command family that turns curated, multi-CLI workflow "matrices" into one-shot, capability-driven installs.
- 30 new CLIs added across creative, productivity, dev, and data domains — including ArcGIS Pro, Calibre, Stata, SiYuan, Joplin, Rekordbox, OpenRefine, Tigris, Mailchimp, MiniMax (chat + TTS)…
- 17 bug fixes hardening existing harnesses — LibreOffice headless conversion (macOS), Zoom recording downloads, Obsidian search, Godot GDScript validation, Browser/DOMShell MCP 2.0.0 migration, n8n…
- Clickable URLs in web fetch and web search tool output
- Context window usage display in tokens at the bottom of the TUI
- Clipboard image paste support in the TUI (macOS)
- Redesigned landing / onboarding for an easier first-run experience
- Faster navigation — improved "go to anything" palette and autofocused search inputs
- Safer deletes — one-click confirmation before removing an app
- Release Notes
- Breaking Changes
- Conversations now load instantly by showing the most recent messages first, then fetching older history as you scroll up.
- You can now test an assistant's connection right from settings, and assistants stay visible even when their backing agent is temporarily offline — instead of silently disappearing, they show a…
- The Skills and Tools settings now read from one unified skill catalog, so what you see in the UI stays consistent with what's actually installed.
- Add session snapshots and revert controls so you can roll a session back to an earlier message, including file changes.
- Always print the MCP OAuth URL so manual sign-in still works when opening the browser flow.
- Add Chrome-style tab cycle shortcuts with mod+1 through mod+9.
- 7d0c73d11f9915e9158063fda299f25eaf2266c5 MUL-3417: tolerate OpenClaw config file CLI mismatch
- b71d9d0ab969cc918b85781d78576af6348b5a7c MUL-3674: Preserve Kiro goal completion on close error
- 5e824a94ae25d1413442ab0123cf017aee69aee1 chore(channel): remove the one-time MULTICA_LARK_HUB_DISABLED cutover switch
- feat(code_match): whole-node boundary \{ P \} ("is")
- ci(deps): bump actions/checkout from 6 to 7
- docs: Hum typography (Plus Jakarta Sans)
- Align code-interpreter with the latest sandbox SDK — Bumped the Kotlin code-interpreter SDK artifact to 1.0.15 and aligned its com.alibaba.opensandbox:sandbox dependency with sandbox SDK 1.0.15.
- @ninan-nn
Jun 24, 2026
- Read command - Added agent-browser read [url] and the matching MCP tool for agent-readable text extraction.
- disable retry for eot errors
- disable eot connection error retry too
- fix(agents): keep STT input anchor on the pipeline across handoff
- fix(desktop): restore solid dot for top-level channel unreads (9d3bfd38b)
- Allow explicit root posts from reply prompts (ebd74d11b)
- feat: publish personas, teams, and managed agents as relay events (d256935c3)
- Updated bundled Claude CLI to version 2.1.191
- Added MCP server instructions to session context. (@Arcadi4)
- Added Opencode-managed provider integration support.
- Added MCP resource template listing.
- Amazon Q for VS Code 2.4.0
- chore(deps): bump the uv group across 32 directories with 3 updates
- chore(deps): bump the uv group across 29 directories with 3 updates
- chore(deps): bump the uv group across 31 directories with 3 updates
- Sidebar: Enabled creating a new worktree from the sidebar's "new thread" icon button.
- Sidebar: Improved the archive "Search all threads" view so you can filter by project name, not just by thread title — handy when many projects are open. (#58214; thanks KevinLaveto)
- Agent: Improved edit tool performance with a more succinct success response.
- Harness no longer exposes a singleton harness.session; callers must use await harness.createSession() and operate on the returned Session.
- Run control moved from Harness.* to Session (e.g. sendMessage, sendSignal, abort, respondToToolSuspension, etc. removed from Harness).
- Event subscription moved from Harness.subscribe() to session.subscribe() (session-scoped event isolation).
- Database migration no longer gets stuck on startup. Two startup migrations could crash and leave the daemon stuck replaying a failing migration on every launch if the database had already been…
- feat: move existing users into the default org when first auto-added
- feat: route unclaimed automation events to the default org on single-org installs
- feat: add hide_personal_workspaces flag for org-only OHE installs
- feat(providers): support profile updates
- feat(sandbox): add GCE metadata emulator for Google Cloud
- feat(sandbox,providers): add aws-bedrock as a recognized inference provider
- Added Knowledge APIs and storage for document creation, listing, patching,
- Added Reflection orchestration for periodically merging and refining episode
- Expanded OME runtime support with event IDs, run-record storage migration,
- Memory (OSS): Guard against malformed image_url entries in parseVisionMessages to prevent crashes
- Memory (OSS): Return attributedTo from get(), search(), and getAll()
- Memory (OSS): Preserve message roles in the extraction input so assistant facts aren't attributed to the user
- Added [update].version_check and [update].manifest_check so background Herdr version checks and remote agent-detection manifest checks can be disabled independently.
- Added HERDR_AGENT=<agent> as a Linux foreground-process hint for agents hidden behind wrappers such as VMs, Bubblewrap, or fence, allowing Herdr to use the named agent's screen manifest when /proc…
- Added ui.pane_borders and ui.pane_gaps to make split pane dividers and spacing configurable.
- Native Responses API transport — Agent turns now prefer the OpenAI Responses API when available, with automatic chat-completions fallback for providers that don't support it.
- Parallel tool calls — A new parallel tool-call runtime lets the agent batch independent tool calls concurrently as visible child chats.
- Revamped MCP server configuration — MCP servers now support project-scoped configuration with global/project merge semantics, a richer settings UI with list-first layout, server and tool search…
- Full Changelog
- b92e4a53fb93f047695cbac9af2453474c1c3dba DH-106 为飞书接入补上 /new 会话指令 (MUL-3503)
- 8ad673fdb7209235f68e2c4c7eff196c265e6ac1 MUL-3560: gate slim runtime brief behind runtime_brief_slim feature flag
- 4d7111d39656bbf9cd30f3d7572a92b6254e6ace MUL-3605: Fix serialization of agent task claims by capacity
- Fixed auth role serialization so request identity roles are consistently serialized as string roles across storage, resource, session, queue, watch, summarizer, and semantic-processing paths.
- Improved user-management setup in the CLI config wizard and stabilized Web Studio account selection.
- Simplified VikingBot auth handling and added trusted auth_mode support for OpenViking-backed bot flows.
- feat: expose configurable websocket max_size limit
- feat: add buffered Chat Completions tool-call streaming
- fix: reduce sandbox sink buffering and spool timeouts
- title: "0.25.1" description: "Smithers 0.25.1 hardens the 0.25 migration path, accounts loading, CLI flag parsing, Gateway auth, Codex subscription preflight, child-failure visibility, and Pi…
- Smithers 0.25.1 is a focused maintenance release for operators and workflow authors upgrading from 0.25.0.
- The release was dogfooded by migrating 12 real stores and checking run-ID parity, inspect parity, fresh compute runs, Gateway reads, and time-travel behavior on the migrated PGlite backend.
- bb desktop 0.0.22
- Adds /fast auto so short conversational calls can start quickly, while longer or fallback work returns to normal mode with the effective state still visible. PR #85104, Issue #85087.
- Shows the effective automatic fast-mode state in status instead of reducing it to on/off, and avoids carrying a cleared Codex service-tier choice into later runs. 8845f2f. Thanks @vincentkoc.
- Keeps automatic fast-mode timing consistent when a turn switches to a fallback model. 075091d. Thanks @vincentkoc.
- The WebUI now sends a browser notification when an agent needs your confirmation or finishes a turn — but only when the tab is in the background, so it never interrupts you while you're actively…
- Fixed the OfficeCLI install link and the outdated German (de-DE) install hint — OfficeCLI is now downloaded as a prebuilt binary from the official releases instead of pointing at the old npm command.
- fix(preview): correct OfficeCLI repo slug casing and de-DE install hint
Jun 23, 2026
- feat(ci): nightly → stable bundles via canonical pre-releases + decision record [gated]
- feat: first-class local-model support for the assistant
- fix: Model handling for tool calling in agents and update IBM models
- fix(release): publish versioned relay Docker tags via independent release lanes (549b7d24)
- fix(desktop): align settings section headers (6ad68a6b)
- fix(desktop): ground agent workspace, migrate legacy nest, configurable repos_dir (1011cea2)
- Changed inherited pi-ai ApiKeyCredential to use the auth.json-compatible discriminator type: "api_key" and provider-scoped env values instead of type: "api-key" and metadata.
- Renamed the inherited agent-core public harness shell execution options type from ExecutionEnvExecOptions to ShellExecOptions.
- Fixed inherited Anthropic-compatible custom models to use explicit compatibility metadata instead of provider-name heuristics for session-affinity headers and unsupported tool-field omissions.
- 1109d0bf64f3d1dd1e1f2bcaf3b38912ec7e833b Get date suffix up to date with release date
- e030fb53ded912d0ce71bd80fe4c0455bc9d5478 .NET: Replace the symlink index entries with regular file entries
- e6ebba1884238134628ad081249b451c0eb15df9 Add ADR 0029: Skills over MCP implementation design options
- Updated bundled Claude CLI to version 2.1.187
- fix: add MAX_REVISIONS guard to prevent infinite revision loop
- Improve SimpleQA eval observability with structured logging and latency metadata
- feat(retrievers): add Brave Search support
- bb desktop 0.0.21
- fix(e2e): align container engine selection across helpers
- fix(python): add encoding=utf-8 to file reads and writes in sandbox.py
- fix(sbom): release lock before sleeping in _rate_limit
- Checkpointing: Added tool batch level checkpointing and unified /continue for regenerate and forking a run. Also added support for forking sessions. See cookbooks.
- Add a new toolkit StudioTool for dynamic agent, team , workflow composition.
- Gemini: GeminiInteractions is now lazily imported, so Gemini no longer requires google-genai 2.0.
- 5038c983c07657d461367970a4c42ab7d8fc9f1c MUL-3281: Add daemon skill bundle refs
- 78342a39ce8697fc7e88fe5e0d04feef0c57a6d1 MUL-3305: feat(agent): add qoder CLI as a choice of agent provider.
- 4ab335b8a5e584e39e15a27034bb5ad72d0ce8f7 MUL-3416: Issue pre-trigger preview + Handoff Note
- Colors + Effects via Metal + Custom Compositor
- Merge pull request #102 from Ti-03/fix/persist-panel-sizes
- refactor(editor): centralize split autosave keys in SplitAutosave
- fix: navigate to new chat after deleting the current session
- feat(console, chat): enhance file preview to support relative path
- fix(queue): bind agent ID at enqueue time to prevent cross-agent deli…
- #1347 e108ba7 Thanks @omeraplak! - Add parallel input guardrails for streamText so async input checks can run while the model starts, buffer streamed output until they pass, and replace blocked…
- Thank you @hliebscher — added the OMLX local provider and helped localize AI routing workload UI.
- Thank you @hgunduzoglu — improved chat layout on short windows, made compatible-provider timeouts configurable, and fixed settings panels calling the correct canonical RPC methods.
- Thank you @ly-wang19 — fixed OpenRouter OAuth callback URLs to respect the loopback listener’s bound port.
- More durable WebUI sessions — Segmented transcript storage keeps large conversations readable without throwing away old turns, fork/replay paths preserve assistant replies more reliably, session…
- Python SDK as a first-class runtime — The SDK now has safer concurrent run() behavior, explicit runtime controls, cleaner per-run hook handling, better MCP cleanup, improved facade lifecycle…
- Gateway and onboarding polish — The gateway gained background/service controls, cleaner foreground shutdown, safer Ctrl+C behavior, and better MCP stdio cleanup.
- Opt-in local daemon for queued writes. mempalace daemon serializes all palace writes through one local process so background work stops racing the palace handle. Opt-in and local-only.
- Opt-in HTTP transport for the MCP server. mempalace-mcp --transport http serves JSON-RPC at POST /mcp (with GET /healthz) for proxy/long-lived deployments. stdio remains the default.
- mempalace_checkpoint batch-save tool. Files multiple add_drawer calls plus an optional diary entry in a single MCP round-trip, verbatim, reusing the idempotent add/dedup path.
- Pass a registered table name (ctx.outputMaybe("research", { nodeId: "research" })) or the schema object (ctx.outputMaybe(outputs.research, ...)) and the result carries that schema's fields, so…
- Drizzle tables resolve through $inferSelect; anything untyped still falls back to the previous untyped row, so existing call sites keep working.
- The CI workflow is green again: every per-package typecheck, the .smithers workflow pack typecheck, the examples and evals typechecks, and the full test suite pass.
Jun 22, 2026
- Local Markdown file links now open inside the AionUi preview panel instead of a browser tab, and support GitHub-style line references (e.g. #L10 or #L10-L20) to jump to a specific line or range.
- Hardened the Windows ARM64 installer: a broken/partial install now fails clearly instead of silently leaving you with a shortcut but no working app.
- Fixed file-change diff previews that could show misleading empty diffs, and added a clearer state when a file can no longer be found.
- Changes since langchain==1.3.10
- release(langchain): 1.3.11 fix(langchain,openai): only set strict=True on tools for OpenAI-compatible models in ProviderStrategy chore: bump pydantic-settings from 2.12.0 to 2.14.2 in…
- build: make mesh-llm opt-in for dev/staging to speed up iteration (ce526106)
- feat(desktop): default temporary channels to 7-day expiry (4b518fc4)
- fix(desktop): render lightbox image sharply at full resolution (ff727af9)
- fix(slng): expose speed in update_options
- fix(google): reject tool calls when tool_choice="none" in realtime
- fix(eot): restore timeout behavior for eot inference
- Step labels appear as the AI reasons — instead of waiting until a step completes, the label streams in with a typewriter effect and a blinking cursor, so you can follow along in real time.
- A dedicated describe panel in the run view — a bordered box below the activity line shows a plain-English description of what the browser just did, updating live as each step finishes.
- The objective header is now a proper bordered box — replaces the flat grey bar, making it easier to visually separate your goal from the step activity below.
- Migrate AWS Sdk to OpenAI
- Fix edit prompt not appearing on user prompt
- add missing headers to ollama embeder engine isAlive method
- Updated bundled Claude CLI to version 2.1.186
- Full Changelog
- fix(multiplexer): stabilize pane lifecycle cleanup
- Show agent provider, model, and variant in TUI
- fix: sync TUI agent variant from message updates
- Enhancement of code match
- fix(code_match): bare keywords, anchored regex matchers, fragment ranges
- test(code_match): cross-check the prefilter on every match test (+ fix it caught)
- MessageQueuePulsar connection retry + graceful drain + subscription cleanup – Added start_with_retry() class method with exponential backoff and optional fallback to FakeMQ.
- Fixed concurrent ReAct cycles on same agent causing LLM API 400 errors – Race condition between NativeHarness phase transition and TaskScheduler _running_tasks cleanup allowed two tasks to run on…
- Fixed completion_timeout leaving zombie tasks that block subsequent rounds – When a round timed out, the underlying task was not cancelled, causing session concurrency slot to remain occupied.
- chore(deps): bump softprops/action-gh-release from 3.0.0 to 3.0.1
- docs(agents): document stale Helm subchart cleanup
- chore(deps): bump actions/checkout from 6.0.3 to 7.0.0
- 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)
- Windows: 下载 GenericAgent-Desktop-Windows-Portable.zip,解压后双击 GenericAgent.exe 启动。卸载时先退出应用,再双击包内 uninstall.bat,最后删除整个解压目录。
- Linux: 下载 GenericAgent-Desktop-Linux-Portable.tar.gz,解压后执行 chmod +x GenericAgent.AppImage,再运行 ./GenericAgent.AppImage。卸载时先退出应用,再运行 ./uninstall.sh,最后删除整个解压目录。
- macOS: 下载 GenericAgent-Desktop-macOS.dmg,双击打开 DMG,把 GenericAgent.app 拖入 Applications,之后在“应用程序 / Applications”中启动。若出现“无法验证开发者”等提示,先双击 DMG 内的 open_anyway.command,再启动。卸载时先退出应用,再从 Applications 删除…
- Snapshot readiness polling — New SandboxManager.waitForSnapshotReady(snapshotId, timeout, pollingInterval) method lets callers block until a snapshot reaches Ready state, replacing hand-rolled…
- replaceContents() empty-body crash — Fixed NullPointerException when calling replaceContents() (verbose=false).
- Added batch multi-file replaceContent e2e test coverage for both verbose and non-verbose paths (Go, Java, C# — matching existing Python/JS coverage).
- Full Changelog
- Diagnostics: write redacted provider reports to a file with platform and app-version context. Thanks @Yuxin-Qiao!
- CLI server: report the startup build version from /health so clients can detect stale helper processes after updates. Thanks @enieuwy!
- Claude: pause background CLI usage probes briefly after rate limiting while keeping manual refresh available. Thanks @kiranmagic7!
- Extension compaction event context - Extension session_before_compact and session_compact events now include reason and willRetry, so extensions can distinguish manual /compact, threshold…
- Safer update flow - pi update installs the exact checked Pi version, and update notices show the changelog URL, making upgrades more predictable. See Install and Manage.
- Added reason and willRetry metadata to extension session_before_compact and session_compact events so extensions can distinguish manual, threshold, and overflow compaction flows
- Merge branch 'main' of ssh://ssh.github.com:443/palmier-io/palmier-pro
- 0.3.6 changelog
- Update CompositionBuilder timing tests
- disable transformers progress bar on Windows (f91529b)
No release notes published.
- cli: add headroom doctor setup diagnostics (e45cf4e)
- cli: add headroom update command and release banner (26be2c3)
- compression extraction — Rust knob exposure, CCR hardening, traffic audits (b7be381)
- TradingAgents v0.3.0 is a stabilization and extensibility release: a verified data-access contract, an expanded provider and data-vendor registry, a current-generation model catalog, and a CI gate.
- Symbol normalization now runs on every vendor path (identity, returns, CLI, news).
- OpenAI-compatible providers register behind a single spec, and a generic openai_compatible endpoint covers vLLM, LM Studio, and relays.
- Browser dashboard. codeburn web serves a local React dashboard with charts and the same task, model, tool, and project breakdowns as the TUI. Reads from disk, binds to localhost.
- Combine usage across your devices. codeburn share and codeburn devices join usage across your machines on the local network, PIN-paired. You can also discover and pair devices from the browser.
- New providers: Grok Build, ZCode / z.ai GLM-5.2, Hermes Agent, Kiro CLI sessions, and zerostack
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