AI agent releases: Jun 8 – Jun 14, 2026
158 stable releases from 72 tracked agents and frameworks in the week of Jun 8, 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 14, 2026
- Windows: .exe installer or portable. macOS: .dmg (Apple Silicon aarch64 or Intel x86_64). Linux: .AppImage (universal) or .deb (Debian/Ubuntu).
- Debian 12+ ships without FUSE 2 by default, which AppImage requires. If ./omniget.AppImage fails with a libfuse error, run
- bash sudo apt install libfuse2
- Documentation-only patch.
- The Codex verification command in docs/codex.md greped for ^codex_hooks\s, a feature-flag name that current Codex no longer prints.
- Fixed (PR
- Plugin client requests now reuse the active server instead of assuming the default local port.
- ACP shell tool calls now show the command and working directory from the start.
- Plugin-provided shell environment variables now apply to PTY sessions.
- Stale sockets are now cleaned up on startup, allowing Crush to run properly
- On unix systems sockets now live in the places they actually belong: $XDG_RUNTIME_DIR on Linux and $TMPDIR on macOS, rather than /tmp.
- 6242e4f47743190fe60a27ba881ba3d6a9b36deb: feat(config): load user-level context files (@Amolith)
- 20/20 templates ship .claude-plugin/plugin.json — minimal plus 19 verticals…
- Per-vertical categories + tags — vertical:devops lands categories: [agent-harness, metaharness-scaffold, vertical, devops/incident-response] so the Claude Code marketplace can filter by vertical.
- 42 new regression tests lock the propagation in — static (every manifest references the entry) + e2e (scaffold proof, substitution works, no leftover {{}}).
- fix(deps): trust sqlite3 so bun's isolated linker builds it ( @hazemadelkhalel
- fix(server): prevent pgvector migration from crash-looping startup on locked-down Postgres
- Added bundled reflect and worktrees skills, including CLI registration, release verification coverage, and documentation.
- Added the /reflect command hook and wired it into plugin command handling.
- Added auto-update syncing for bundled skills so newly shipped built-in skills can be installed after plugin updates without overwriting existing user skill folders.
- Install with
- npm install -g codeburn codeburn menubar
- That command drops the app into ~/Applications, records the persistent codeburn CLI path used by the menubar, verifies the downloaded checksum, clears quarantine after bundle verification, and…
- 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)
- Default URL in config — set a default start URL with config set-url and the CLI will use it for every run automatically.
- Per-run --url flag — pass --url <address> on any kane run or testmd run call to override the default for that run without touching your config.
- URL in .testmd frontmatter — add a url: key to a test file's frontmatter and it becomes that file's built-in start URL. Resolution order: --url flag → frontmatter → stored default.
- fix(ci): trigger generate-cli-docs hook on cli.mdx edits
- Agent experience: .md mirrors, docs llms.txt/llms-full.txt, skill endpoint, examples agent docs
- Agent experience follow-ups: fix review findings from
- fix(release): publish manifest from successful platforms (9b410325a)
Jun 13, 2026
- Improved MCP server compatibility by declaring OpenCode's supported client capabilities.
- Memory: Add a contextual OSS-to-Platform notices system that surfaces occasional, situation-aware messages…
- Dependencies: Upgrade @langchain/community to ^1.1.18 to remediate CVE-2026-27795 and CVE-2026-26019
- Dependencies: Resolve all open MEDIUM Dependabot alerts via pnpm overrides
- Full Changelog
- Fixed inherited OpenAI GPT-5.4/GPT-5.5 and OpenAI Codex GPT-5.4/GPT-5.4 mini/GPT-5.5 context window metadata to use the observed 272k-token Codex backend limit, avoiding a billing hazard from…
- Codex Stop hook no longer blocks a normal stop on an incomplete plan (PR
- SHA-cache documentation corrected to the v3 location (PR
- Native Codex PreCompact hook (PR
- 慢 provider 和桌面长任务更稳定:桌面端不再用过短的超时杀掉健康但首字节很慢的流式请求,非流式 fallback 状态会显示在聊天状态栏,被中止的 API call 也会进入 Trace 而不是一直 pending,重点跟进 #766 和 #773。
- 回答进行中可以排队输入:active turn 运行时支持继续提交、引导、编辑和删除排队消息;CLI 接收后自动回放进 transcript,并避免 guided prompt 或 replay 造成重复用户气泡,修复 #755。
- H5 远程访问稳定性升级:H5 token 和端口可跨重启保持,断连宽限时间可配置,手机锁屏或切后台时不再强杀正在运行的 CLI;触屏 H5 修复输入框缩放、键盘遮挡、长按选择、操作按钮不可见和安全区适配问题,跟进 #764、#767 和 #780。
- fix(release): resolve Windows sidecar path and Linux AppImage updater format (c7dd4295b)
No release notes published.
- bb desktop 0.0.18
- 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 12, 2026
- Sign in through your browser. Logging in now hands you off to your normal browser (Safari, Chrome, whatever you use) and brings you back.
- See how long things take. In-progress steps show a live timer so you know it's working. Hover over its "thinking" to see how long it spent reasoning. Less guessing about whether it's stuck.
- Find within a conversation. Search inside a long chat to jump to a specific word or moment.
- Fixed Windows ARM64 installs by falling back to the Windows x64 binary during postinstall, avoiding failed downloads for a native ARM64 artifact that is not published
- @EternalRights
- [codex] add BU3 browser use models
- [ENG-4762] publish: gate publish_to_pypi behind release env
- feat(llm): accept provider-prefixed models in ChatBrowserUse
- This release brings a faster, more transparent desktop experience—more concurrency in chat, smoother reliability in memory flows, and easier access to diagnostics. 🚀
- Thanks to everyone who contributed to this release: @github-actions[bot], @sanil-23, @senamakel, and @senamakel-droid.
- Clearer Bedrock validation guidance - Amazon Bedrock data retention validation errors now link to AWS data retention documentation. See Amazon Bedrock.
- Added an experimental first-time setup flow behind PI_EXPERIMENTAL=1 that asks for a dark/light theme choice (preselecting the detected appearance) and opt-in analytics data sharing on first…
- Added AWS data retention documentation links to inherited Amazon Bedrock unsupported data retention mode validation errors
- feat: add zellij current tab pane mode
- feat: map zellij main layouts to pane directions
- [BETA] - Version 2
- Clipboard, cookie, and localStorage are now first-class tools — test flows can read and write the virtual clipboard, and can perform full create/read/update/delete operations on cookies and…
- A live hints bar at the bottom of the TUI — a new footer row shows tips relevant to your current mode, cycling through a remote catalog that updates automatically (cached for 1 hour).
- Hints can be turned off — run /config inside the TUI to toggle hints on or off; the setting is saved locally and defaults to on.
- SkillPlugin – Give agents access to local skills aligned with the agentskills.io spec. Skills become a first-class, portable extension surface for AG2 Agents.
- Amazon Bedrock Beta Client – New Beta client for Amazon Bedrock, bringing Bedrock into the Beta config architecture.
- Decoupled UsageEvent – New usage reporting via a decoupled UsageEvent for cleaner separation between LLM calls and usage tracking.
- release(langgraph): 1.2.5
- fix(langgraph): merge lc_versions config metadata
- release(cli): 0.4.28
- bb desktop 0.0.17
- Speech now transcribes live as you talk — partial and final text appears in the input box in real time, in both the chat send box and the home page…
- macOS fix: the microphone now works under the hardened runtime. Previously Speech-to-Text silently recorded nothing and the app never appeared under Privacy & Security → Microphone.
- Assistant Settings is now a full-page governance area: builtin and custom assistants live in one list, with a full-page editor replacing the old drawer, and assistant-bound conversations are…
- Changes since langchain==1.3.8
- release(anthropic): 1.4.6 release(langchain): 1.3.9 fix(langchain,anthropic): confine file-search results and tighten anthropic allowed_prefixes
- #1656 4c2d1a7 Thanks @cjol! - Rebuild agents/browser on the codemode connector runtime (experimental).
- BrowserConnector — a CodemodeConnector (name cdp) that owns CDP sockets keyed by execution id. Sockets are released at the end of every execution pass (onPassEnd).
- Session modes — one-shot (default, fresh session per execution), reuse (named shared session), and dynamic (starts one-shot; the model can promote with cdp.startSession() after e.g. logging in).
- Learnings CRUD on AgentOS: Added create, read, update, and delete endpoints for learnings on AgentOS.
- Gemini Thread Safety: Removed per-response Gemini cleanup that caused a thread-safety race under concurrent usage.
- JSON Object Providers: JSON instructions are now included in the followup prompt for json_object providers.
- Full Changelog
- docs: changelog and blog post for v0.8.1
- chore(integrations): apply CI ruff formatting to haystack + roo-code tests
- blog: Cline persistent memory (lifecycle hooks, no MCP)
- fix(server): use public tonic body type in gRPC rate limiter
- feat(gateway): add reconciler lease for HA multi-replica deployments
- feat(gateway): add system registry support and source indicators
- Removed harness.respondToQuestion(...) → use harness.respondToToolSuspension(...).
- Removed ask_question event → listen for tool_suspended and read event.suspendPayload.
- Removed Harness question-related APIs/types (registerQuestion, pendingQuestion, HarnessQuestion* types); pendingSuspension becomes pendingSuspensions: Map<toolCallId, ...>.
- Installs now pinned + integrity-checked
- fix(docs): escape user input in the demo terminal
- Real lifecycle hooks: event dispatcher for session init, chat.message
- [Experimental] before_tool and after_tool hooks: shell scripts declared in hooks.toml that fire around every tool call.
- Message queue: messages typed while the agent or a !bash command is running are queued and shown above the input.
- Tool result output is now collapsed by default to keep responses scannable
- Mitmproxy static options moved from hardcoded flags to config.yaml — all static mitmproxy options (mode, listen_host, stream_large_bodies, ssl_verify_upstream_trusted_confdir, ignore_hosts) are…
- OPENSANDBOX_EGRESS_MITMPROXY_CONFDIR — removed because it would break config.yaml discovery.
- OPENSANDBOX_EGRESS_MITMPROXY_IGNORE_HOSTS — replaced by the yaml-native list in config.yaml (the env-based path was the root cause of the silent-overwrite bug).
- Windows: .exe installer or portable. macOS: .dmg (Apple Silicon aarch64 or Intel x86_64). Linux: .AppImage (universal) or .deb (Debian/Ubuntu).
- Debian 12+ ships without FUSE 2 by default, which AppImage requires. If ./omniget.AppImage fails with a libfuse error, run
- bash sudo apt install libfuse2
- fix(memos-local-plugin): guard OpenClaw runtime startup
- fix:The failed task was wrongly recorded as a "successful experience".
- docs: add Chinese translations for evaluation documentation
- add differential network capture harness (11ab5f8)
- add light mode for dashboard (c425893)
- add OAuth2 client-credentials upstream-auth proxy extension (eb2e50f)
- wren-core: RLAC subqueries on MDL models + to-many calc on non-PK column (6972e69)
- fix(py): make coco.fn directly callable outside a component context
- refactor(state): move UserState db key from 0xc0 to 0x34
- docs: add landing / About page
- Multi-agent runtime: run any number of named agents from one daemon, each with isolated workspace, memory, skills, model provider, and risk profile.
- Schema V3 with automatic migration: existing configs migrate in place on first load.
- zerocode: a new terminal UI with dedicated Chat, Code, Dashboard, Config, Quickstart, and Logs panes.
Jun 11, 2026
- test: categorize tests and select via pytest --unit/--plugin openai/ … flags
- hamming plugin: add customer_metadata, external_links, and session_mode fields
- fix(job): log unhandled exceptions in fire-and-forget tasks
- fix: kill local process group on timeout
- @Swing-G made their first contribution in
- maaz139 for reporting the timeout bug
- macOS arm64 companion binary for oh-my-opencode-slim.
- fix(SKY-11177): harden element-tree scrape against undefined className and isElementVisible recursion (ada154bd1)
- fix: stop the docker UI from filtering out its own injected API key (403 "Invalid credentials" on self-host)
- SKY-10827: persist copilot artifact-metadata violation batches as durable telemetry
- #1710 df76207 Thanks @seanmcguire12! - handle target closed errors on rapidly opening/closing popups
- #1593 159f6d3 Thanks @tkattkat! - Update extract tool
- #1396 b7bc0b6 Thanks @seanmcguire12! - deprecate drawOverlay parameter in observe
- MiniMax M3 added as an option.
- Web searches are easier to follow. When the assistant looks something up, results appear as clean, clickable cards in the conversation. See what it searched and click straight to the sources.
- New geo-audit skill for location and SEO analysis, now in the catalog.
- Amazon Q for VS Code 2.3.0
- [fix] export text first frame empty
- theme color
- Materialize preview text layers lazily around the playhead
- Add pluggable default backends for memory, knowledge, rag, and flow.
- Surface real finish_reason, sampling params, and response.id on LLM events.
- Type DSL triggers as route-aware decorators.
- fix(stt): send multipart request matching backend /api/stt contract
- ci(build): support AionCore manual artifacts
- feat(theme): add Follow System theme mode to gallery
- chore(release): stamp v0.1.5 [skip ci] (e5fdd95)
- fix: zombie processes, atomic writes, CI --release v0.1.5 (5a0b501)
- Make devcontainers working (e29d043)
- Model providers
- Implements auto-populated model lists for multiple providers, eliminating the need to type model names manually.
- Allows configuring multiple API keys for the same model provider. #14595
- style: truncate tool card titles to single line with ellipsis
- chore: bump version to 1.1.11post2
- Lifecycle-authority agent integrations such as Pi and OpenCode no longer trigger a repeated detection reset loop that could flood logs, drive high CPU, and make the UI lag or stop responding.
- 客户端/服务端架构拆分:记忆引擎从 OpenClaw 嵌入式插件拆分为独立 Gateway 服务进程,部署方式与接入方式发生变化。
- 配置结构变更:插件配置结构扁平化重构,原 gateway 字段迁移为 server 嵌套。
- 插件入口模式变更:支持 local(进程内本地运行,默认)和 client(连接外部 Memory Gateway)两种接入模式。
- 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)
- ROB-3942 Hardwire cluster and conversation context into platform-mcp headers
- override config file
- Add Naomi Caren to MAINTAINERS.md
- fix: expose sandbox error retryability
- fix: #3512 type tool-end hook results as object
- fix: use tuple form for SpeechGroupSpanData __slots__
- Migrate code-interpreter paths from /opt/opensandbox to /opt/code-interpreter — all code-interpreter entrypoint, env script, and log paths have moved from /opt/opensandbox/ to /opt/code-interpreter/.
- Sign and attest release artifacts — code-interpreter Docker images are now signed with Sigstore/cosign keyless signing and published with GitHub attestations.
- Add seccomp-based clone3 fallback in execd — inside the sandbox, execd now falls back from clone3 to clone when the seccomp profile blocks clone3, improving compatibility with older kernels and…
- coco.use_state() — persistent per-component user-defined state @hakuuww
- LiveMap — in-memory intermediate collection @georgeh0
- Preview mode for update actions @MrAnayDongre
Jun 10, 2026
- Change(swe-bench configs): Drop T=0 default in swe-bench
- Enh: Expose finish_reason to format_error_template
- feat(inspector): display reasoning_content with toggle and hot-reload
- Sub-Agent Event Streaming: Sub-agent events from the context provider update tool now stream through to the parent run.
- AgentOS Registry Auto-Population: The AgentOS registry now auto-populates from agents, teams, and workflows.
- Workflows HITL over Sockets: Added socket support for human-in-the-loop workflows.
- Copy mode page scrolling now stops at the same top and bottom boundaries as normal pane scrolling instead of overshooting or getting stuck near the edges.
- Clipboard-copy feedback no longer stays visible after the related selection state has gone stale.
- The session navigator now uses live workspace labels, so renamed workspaces and cwd-derived labels stay current while navigating.
- Bump version to 1.0.40 (def90269c)
- revert(SKY-10083): revert screenshot downscaling — no measurable token savings (ea9907cf0)
- refactor(browser-profiles): single source-of-truth for paginated row selection (SKY-10853) (6f7a5f5bf)
- Agent: Added a way to share skills via links.
- Agent: Added support for symlinked global skill directories. (#58095; thanks David3u)
- Agent: Added a manual way to trigger the Rules to Skills migration.
- Local files in generate mode — pass --files on the command line or type @filename inline to attach files to a generation request.
- @-mention selector with grouped categories — typing @ opens a unified palette that organizes matches into Files, Scenarios, and Test Cases, with a bounded 7-row scroll window so it never takes…
- Mistyped @ paths are surfaced, not silently dropped — if a referenced file can't be found, kane-cli warns you instead of ignoring it.
- Changes since langchain==1.3.6
- release(langchain): 1.3.7 style(langchain): add ruff rules ARG feat(langchain): add ProviderToolSearchMiddleware chore(langchain): activate mypy warn_return_any test(langchain): mark legacy…
- Session Trace Monitoring:新增会话 Trace 记录和查看能力,支持按 session 查看 LLM 请求、工具调用、消息详情、耗时与错误信息;Trace 界面升级为更接近 LangSmith 的双栏树状布局,并把入口收敛到 Settings。
- IM 适配器增强:新增 WhatsApp linked-device 支持,完善配对、媒体处理和桌面设置入口;Telegram 新增命令菜单,常用操作不再依赖手写命令。
- 桌面多语言补齐:新增日语、韩语、繁体中文界面,并补齐后续新增设置项、Trace、Adapter 等页面文案。
- dd29f9aa654257ad807ef7fdcd25fcc7ded109f6 .NET: Hosted Agent Sample - Toolbox with various Auth
- a5f4e0078e1c0d1d8fcd580a1bcccc8836093b98 .NET: Fix .NET Copilot integration tests for SDK v1.0.0
- 3c0c12cd463601c451130884b5442dc42b12cf1e .NET: Update release version for 2026-06-10 release and switch GH.CP Agent to RC
- Help Center search and per-locale portal branding.
- WhatsApp voice messages and improved Twilio call controls.
- Processing status for one-off SMS and WhatsApp campaigns.
- Fixed click reliability by scrolling off-viewport elements into view before resolving coordinates, handling JavaScript dialogs promptly, recovering mouse state after dialog-opening clicks, and…
- Fixed frame-scoped selectors and waits so CSS-selector actions and wait respect the selected iframe, including cross-process iframes, with translated click coordinates
- Fixed wait timeout handling so the documented 25s default is used, --timeout is honored across wait variants, and long waits receive an appropriate client read budget
- Added LLM profiles
- Users can select the sandbox grouping strategy
- Enabled sub-agent delegation
- fix(ci): handle empty release prefix check
- fix(http): read error response body once to avoid double consumption
- fix(preview): point OfficeCLI install help to official releases
- npm: bash npm install -g base44@latest
- Homebrew: bash brew install base44/tap/base44
- Version: 0.0.55 NPM Tag: latest
- Free Model OAuth: Zero-config free models with one-click OAuth authentication
- Xiaomi MiMo Provider: Add Xiaomi MiMo Token Plan as built-in provider
- Extra Body Routing: Route non-standard generate_kwargs into extra_body
No release notes published.
- 351bf79fcec913a762f22334062bf22af97fbc45
- 006945cdc227f743d0308704c27e259ed677d64f
- New Breaking Changes for terminal
- 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)
- Full Changelog
- Knowledge Base: Fixed the issue where the QA pair template content was displayed incorrectly when downloaded.
- Agent: Fixed the issue where questions could not be sent during debugging for agents with user input parameters.
- Agent: Fixed the required field error when directly asking a question with the default value kept for an external toggle parameter whose default value is false.
- v3 targets long running agentic runs on strong models (Opus 4.8, Fable 5, GPT 5.5 class). Everything new is opt-in.
- init-session.sh --autonomous keeps the turn start plan injection and drops the per tool call re-injection that the v2.21 eval measured as a 68 percent token tax.
- init-session.sh --gated adds a deliberate completion gate on the Stop hook. The gate blocks a stop only when all five conditions hold: the plan opted into gated mode, a phase is in_progress…
- live: Handle input transcription differently for Gemini Live 3.1 models (ecfdaf5)
- add missing Gemini imports in base_llm_flow (6d027b4)
- flows: Reset reconnect attempts on connection success (87abf23)
- [tests] refactor into subfolders
- Add remove_tracks tool, transcript paging cursor, and get_timeline windowing
- Allow zooming out to fit long timelines; log-map the zoom slider
- 261 public skills across coding, research, security, media, enterprise ops, and agent workflows (rc.1 shipped 243) — plus 64 agents and 84 commands.
- The control-pane substrate — the foundation of the ECC 2.0 operating surface
- harness-neutral session adapters (ecc.session.v1) covering Claude Code, Codex, OpenCode, and dmux — one schema for "which agent is where, doing what"
- support claude fable 5
- anthropic: tool choice auto when thinking (cannot force it)
- Fix packaged agent tools env
- Add Claude Fable 5 to the model selector
Jun 9, 2026
- Install with
- npm install -g codeburn codeburn menubar
- That command drops the app into ~/Applications, records the persistent codeburn CLI path used by the menubar, verifies the downloaded checksum, clears quarantine after bundle verification, and…
- We did a minor patch just for this one thing
- Claude Fable 5 is now available as a model option. Only to remove it in a later build due to this
- agent-framework-core: Add MCP client OTel spans per GenAI semantic conventions
- agent-framework-core: Add MCP long-running task support
- agent-framework-claude: Bump claude-agent-sdk to 0.2.87
- Company Artifacts - Files, media, and documents your agents produce are now first-class.
- Collapsible sidebar rail and takeover panes - The primary navigation can now collapse to a persisted rail with hover/focus peek, giving contextual pages — especially company settings and plugin…
- Rich issue attachments with video - Issues now accept video attachments and render rich inline previews, including standalone PWA browser controls and inline video playback, so screen recordings…
- fix(xlsx): shift embedded formulas in dataValidation/conditionalFormatting/table on row insert/delete
- fix(xlsx): clamp conditional-format sqref expansion to the sheet's used extent
- @dragonwhites made their first contribution in
- Fix(swebench): source ~/.bashrc via BASH_ENV to activate testbed env
- fix: persist full model response in extra dict on FormatError
- @VittoriaLanzo made their first contribution in
- Claude Fable 5 - Claude Fable 5 is now available on the Anthropic and Amazon Bedrock providers, with adaptive thinking and xhigh effort support.
- Prompt template defaults - Prompt templates can use default positional arguments such as ${1:-7} for optional values. See Prompt Template Arguments.
- Configurable project trust defaults - defaultProjectTrust lets users choose whether unresolved project trust asks, always trusts, or never trusts by default, and extensions can inspect effective…
- docs: add Anthropic Managed Agents browser guide
- [STG-2188] docs(cli): document full browse command surface + fix license badge
- [chore]: use allowlist for published files
- test(cli): avoid browser launch in auth rollback test
- docs: document DCO commit sign-off requirement
- refactor: deduplicate OCSF builder setters and persistence helpers
- feat: allow users to name flow snapshots when saving (LE-874)
- feat: Ship Calculator default tool + dynamic system prompt injection
- feat: disable frontend retries for client errors
- Make TUI quit instant by SIGKILL-ing the sandbox container
- Swallow sandbox container races in the stream consumer
- Strip all images from session on vision-rejection, not just the latest
- New conversations now use the ACP and AionRS runtimes; older Codex and legacy conversations remain readable but read-only.
- OpenClaw now appears as an ACP backend when its local command is available.
- The conversation view is now turn-aware, so late send, stop, hydrate, and completion events can no longer overwrite a newer or already-completed turn — fewer stuck or out-of-order UI states.
- [x] Cerebres provider
- [x] The default chat thread is now killed when you create a new thread. If you have chats on the default thread, it will be available still.
- [x] All model providers are now opt-out of tool calling by default. Everything will call tools by default unless you opt-out offering better performance for agents everywhere
- AzureOpenAIChatGenerator now accepts a Secret for the azure_endpoint and api_version parameters in addition to a plain string.
- fix(opencode): fold recall into the first system section, not a new one
- docs: changelog and blog post for v0.8.0
- feat(roo-code): package as installable PyPI CLI (hindsight-roo-code)
- feat: data manipulation triggers @Louai-Zokerburg
- Automatic review after idle: Once a conversation goes idle, the agent reviews it in the background to fix problems a skill exposed in use, create reusable new skills, follow up on unfinished…
- Silent by default, notify on demand: It reports what it changed only when it actually made a change, and stays silent otherwise
- Safe and reversible: Every review is backed up beforehand and can be undone at any time. Built-in skills are protected, and all reads and writes stay within the workspace
- 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)
- Update audio agent guidance for Lyria
- [fix] Don't color-tag ProRes 4444 writers (zeroes alpha)
- [fix] tool json output capped to 2 decimal places for float
- perf: add --format {text,json,csv} to headroom perf (9fe4886)
- proxy: show resolved upstream API targets in startup banner (8dbe7ad), closes
- relevance: weight BM25 score_batch by corpus IDF (88177bd)
Jun 8, 2026
- createSmithersPostgres boots node-postgres or an embedded PGlite. A new async API stands up the engine against a real Postgres server or an in-process PGlite over a local socket, so you can…
- The dialect is exercised end to end on real PGlite. New test suites boot an in-process PGlite and run the db dialect, the engine builder, and the time-travel fork/snapshot paths against it…
- A serial snapshot service per worktree. createSnapshotService (packages/engine/src/snapshotService.js) runs a per-cwd serial queue so jj never races itself.
- add x402 skill
- Magnus/rust sdk server clean pr
- Validate Rust terminal SDK server binary
- Bare-objective shortcut removed. kane-cli "<objective>" no longer routes to run. Use kane-cli run "<objective>".
- Exit code 1 → 2 for TMS credential-exchange failures. Aligns with the other auth/setup failure codes. CI scripts that
- config show and the new list commands emit JSON when stdout is piped or redirected (> file, | jq) instead of the human table. Useful for scripts; anyone scraping the human format will need to update.
- #1701 6caa6e8 Thanks @mattzcarey! - Refactor WorkerTransport to extend the official MCP SDK's WebStandardStreamableHTTPServerTransport instead of being a hand-rolled implementation.
- CORS — preflight handling and response-header injection (corsOptions).
- Persistent transport state across DO hibernation via the existing MCPStorageApi adapter.
- #1337 685f8e4 Thanks @omeraplak! - Sanitize tool call inputs before model replay so malformed or non-object values cannot break provider history conversion.
- If you accidentally installed the Intel (x64) build on an Apple Silicon Mac, AionUi now detects this at startup and shows a clear prompt to download the correct build — instead of failing in a…
- More reliable managed-runtime preparation: hardened ACP bundle setup, validation via real entrypoints, and clearer failure details when resource preparation goes wrong.
- ACP now prefers configured option catalogs and handles the Hermes "yolo" fallback correctly.
- refactor(tui): extract shared setting edit overlay
- feat(kubernetes): support driver config passthrough
- fix(sandbox): stop log push after auth failure
- Project trust for local inputs - Pi now asks before loading project-local settings, resources, instructions, and packages, with saved decisions and --approve / --no-approve controls for…
- Extension-controlled trust decisions - Global and CLI extensions can handle project_trust, decide, remember, or defer project trust before project-local resources load. See project_trust.
- Cache-hit visibility in the footer - The interactive footer now shows the latest prompt cache hit rate (CH). See Interactive Mode.
- > CodeWhale is the canonical project, command, npm package, and release-asset name. The legacy npm package deepseek-tui is deprecated and receives no further releases.
- bash cargo install codewhale-cli codewhale-tui --locked
- Both crates are required: codewhale-cli provides the codewhale dispatcher and codewhale-tui provides the interactive runtime.
- fix(deps): resolve all open Dependabot alerts in uv.lock files
- fix(falkor): strip nul bytes from parameters
- fix: FalkorDB profile failing to start in docker-compose
- docs(guides): add Hermes memory guide batch
- docs: changelog and blog post for v0.7.2
- feat: upgrade MiniMax default model to M3
- /teleport slash command exposed over ACP, mirroring the TUI command for IDE integrations
- Startup prompt to install a pending Vibe update before continuing the session
- JetBrains IDEs no longer trigger a preemptive auth prompt over ACP and no longer drop terminal arguments
- Windows:下载 .exe 文件,双击运行安装向导即可。如遇 DLL 缺失,请安装 VC++ 运行库。
- macOS:下载 .dmg 文件,打开后将应用拖入「应用程序」。首次打开如遇安全提示,前往「系统设置 → 隐私与安全性」允许运行。
- Linux:下载 .AppImage,执行 chmod +x ToonFlow-*.AppImage 后运行。
- Remove agent-browser skill install
- Feature: Env support in config
- fix(hooks): prevent phase-reminder from leaking into UI/chat history
- fix: re-inject UI env vars on every container start (7f97fe8fd)
- fix(SKY-9402): skip force-visible for shadow DOM inputs in closed dropdowns (79dfc8aa3)
- perf(SKY-10319): add screenshots_included/speculative attrs to LLM request spans (1d1076ab5)
- Bumped the Kotlin Code Interpreter package to 1.0.12 and aligned its Kotlin Sandbox SDK dependency to 1.0.12, matching the latest sandbox release train.
- @ninan-nn
- Full Changelog
- Models: Fixed the error when adding reranking models from Alibaba Cloud Bailian.
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