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: Jul 13 – Jul 19, 2026

222 stable releases from 83 tracked agents and frameworks in the week of Jul 13, 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.

Jul 19, 2026

  • Unicity AOS Community Edition 2026.1.2.
  • This product release bundles Astrid Runtime v0.10.2. The exact runtime release identity, WIT commit, and compatibility pins are published in runtime-compatibility.toml and in every archive's…
  • The release also contains the 19 signed, installable capsule artifacts selected by Community Edition. Published astrid-capsule-* identities remain unchanged.
No release notes published.
CodexBar
  • Refresh: prevent macOS 14 launch crashes caused by TaskLocal task-allocation corruption (#2341, fixes #2319 and #2326). Thanks @lzylzylzy130 and @jorgesancha!
  • Menu bar: render custom-layout provider icons at the native size and tint them for light and dark menu bars. Thanks @elpinguinofrio!
  • Menu: fix switcher “Weekly progress” to prefer weekly quota windows, with provider-specific fallback when unavailable. Thanks @Anneo22!
  • Polish onboarding runtime error states (31090b186)
  • feat(desktop): add tooltips to channel header actions (cab47905a)
  • feat(desktop): add focused thread mode (2d0bd6ec4)
  • Added HAR response body capture with text bodies embedded by default and configurable all, text, and none content modes across the CLI and MCP surfaces
  • Added a derive-client skill for recording browser traffic and generating reusable API clients from HAR request and response data
  • Open the workbench in a browser. The desktop backend can optionally serve the same renderer to a browser on your own computer, bound only to 127.0.0.1 behind an authenticated URL, and can run…
  • Drive it from the command line. A bundled open-science CLI manages the headless backend as a background daemon — start, status, url, stop — so the browser is the only UI and closing your terminal…
  • Manage the agent backend yourself. Settings cards now let you install, switch, and uninstall the app-managed Claude Code and OpenCode runtimes.
  • fix(ci): avoid protected branch push in plugin releases
  • release: @memtensor/memos-local-plugin v2.0.10
  • Fix #2090: fix: README benchmark scores in News section don't match Performance table (LoCo
  • Closed cross-agent memory leak via /agentmemory/context under AGENTMEMORY_AGENT_SCOPE=isolated
  • Hooks no longer crash ("hook exited with code 1") on a null payload from Codex CLI / Claude Code
  • Engine conflict no longer hangs in a WS reconnect loop; clean recovery and pinned-engine guidance
  • npm: bash npm install -g base44@latest
  • Homebrew: bash brew install base44/tap/base44
  • Version: 0.1.5 NPM Tag: latest
  • 0.6.12 changelog
  • [perf] Reduce audio meter rendering work
  • [mcpb 2.0] Survive Palmier restarts on the Claude Desktop MCP connector
  • Full Changelog
  • Updated bundled Claude CLI to version 2.1.215

Jul 18, 2026

Buzz
  • fix(desktop): recover first community joins (775cf672d)
  • fix: recover community access visibility (ca384d082)
  • 996c83bdeeea3a2eb5e37a6a0446b861b9f555f5 Merge pull request #608 from pinchtab/feat/interactive-annotate
  • 42b81ce41ce7e6486108c8d3eaf446f4cf146acc Merge pull request #610 from pinchtab/feat/site-scrape
  • 958ecad00c734c3f30fb23923503ddc800b2444a build: consume released seaportal v0.2.0
  • Menu bar: add drag-and-drop layouts with customizable identity, usage, reset, cost, spacing, and stacked-line tokens.
  • Usage: estimate weekly quota in full 5-hour windows and show whether it can run out before reset. Thanks @hdsheena!
  • CLI: add quota-aware codexbar guard automation gates with stable exit codes, explicit windows, JSON output, and bounded fetches. Thanks @OfficialAbhinavSingh!
  • chore: upgrade dependencies (TypeScript 7, jsdom 29, biome 2.5, zod 4.4, etc.)
  • fix: quiet repeated skill update notices
  • Make background job board injection cache-safe
  • Fixed the titlebar minimize / maximize / close buttons doing nothing on Windows and Linux (a v2.1.36 regression).
  • The team member "processing" timer no longer restarts from 0s when you open settings and come back, toggle fullscreen, or resize the window — it now shows the true elapsed time.
  • fix(bridge): accept void-param invokes after JSON serialization
  • Choose your agent backend (Claude Code ↔ OpenCode). The ACP runtime is now generic and delegates every framework-coupled decision — spawn, model config, session setup, permission mapping, MCP…
  • An opt-in reviewer that checks the agent's work. After a turn completes, an optional reviewer runs in a clean, tool-restricted context and traces the agent's claims against the real transcript…
  • Stays out of the way, and cleans up after itself. A system-tray presence with single-instance locking keeps exactly one running copy, and quitting now awaits cleanup of the agent child process so…
  • 会话恢复与撤销更可信:撤销当前轮次改用真实 checkpoint;重启前会先落盘上下文,恢复历史会话时不再误触发预热,长对话重新挂载后保持阅读位置。
  • 后台任务终态持久化:后台 Agent 的完成、失败和停止状态会写入会话历史;完全退出再启动后,已完成卡片不会回退成“进行中”,旧版通知格式仍可继续读取。
  • H5 断连清理补齐启动竞态:手机在 CLI 启动阶段离开、权限请求随后才出现时,服务端仍会建立清理观察;无人连接的权限等待会在上限后停止,重连和正常长任务不会被误杀。
  • Help Center improvements including staged edits for published articles, admin search, article reordering, and popular content by locale
  • Captain assistant overview and drill-down analytics (Enterprise)
  • Conversation and message drilldowns for reports
Planning With Files
  • plan-doctor.sh (new in v3.6.0) now ships in every synced IDE skill folder, not only the two canonical scripts/ locations.
  • docs/gemini.md routes new installs through the standard .agents/skills/ path, which is version-locked, instead of the intentionally version-lagged .gemini/skills/ variant…
  • AGENTS.md and the bump-version.py docstring now describe the actual 18-entry parity set.
  • security: sanitize untrusted content, add SECURITY.md, pin brotli
  • Fix silently dropped token limits, real usage-based cost tracking, and research pipeline robustness
  • Add Deep Agents example: GPT Researcher as the research engine in a LangChain deep agent
  • livekit-plugins-fishaudio: pooled websocket reuse + prebuffer to fix cold-start crackle
  • rename reload-addr flag to cli-addr
  • fix(drive-thru): instruct agent to add stated quantities without confirming
  • feat: add cooperative timeout support
  • feat(py): RetryWithSmallerBatch — split failed batches, retry halves
  • fix: commit dbi-opening read txns to keep LMDB handles valid across processes
No release notes published.
  • Add include_model_request_parameters instrumentation setting to omit the model_request_parameters span attribute
  • Add RaiseContentFilterError capability to raise content filter error for non-empty content filter responses
  • feat(usage): add cache_hit_ratio property to RequestUsage and RunUsage
  • Updated bundled Claude CLI to version 2.1.214
  • bb desktop 0.0.32
  • Added support for stacking multiple leading slash commands in a single prompt, so a chain like /plan /grill-me <prompt> parses, activates, and renders every command in the order you typed them.
  • Improved scrolling in the /diff viewer so paging through a diff no longer jitters or pushes the status line off the screen when lines wrap or comments expand.
  • Fixed custom agents that declare subagent: false still appearing in the available-subagents list and being invocable as subagents.
  • Keyboard-first navigation — arrow keys, Enter, and digit shortcuts move through the panel; typing immediately opens a free-text editor row seeded with context so you never start from a blank prompt.
  • Cleaner visual hierarchy — work output indents by two columns, receipts are machine-only, and human-facing commit lines are kept to a single line.
  • Colored risk and recommended-row highlighting — the question panel announces its header, marks risk level in color, and highlights the recommended answer so the right choice is obvious at a glance.
  • Move Elixir observability dashboard to Phoenix
  • Refactor config access around an Ecto schema
  • Align workflow config spec with schema behavior

Jul 17, 2026

  • bb desktop 0.0.31
Buzz
  • fix(desktop): honor selected onboarding runtime config (0bbcafe9e)
  • Polish community onboarding flow (2ea71de24)
  • fix(desktop): retry transient runtime installs and repoint Goose to official upstream (51c2d1ca6)
  • Updated eve extension packaging for eve 0.25.1, adopting the new source and dist extension manifest format, updating the eve example to stable AI SDK releases, and keeping extension tests aligned…
  • @AndrewBarba
Open Interpreter
No release notes published.
CodeBurn
  • Windows: fixed the bundled-CLI 'not found' that broke every Windows install (path resolution).
  • Unified solid-flame brand across the app, dock, and website.
  • Richer error diagnostics in anonymous telemetry.
OfficeCLI
  • Full Changelog
  • [fix] Serialize MLX inference behind a process-wide gate
  • Move export summary scan off main actor
  • capture export failure diagnostics
  • Add rubric-based partial-credit scoring and contradiction detection
  • Add reproducible sampling, quick runs, configurable models, concurrency, and timeouts
  • Improve eval reporting with metadata, latency, and failure metrics
  • fix(pieces): file uploads no longer send empty or corrupted files @majewskibartosz
  • fix(approval): Create Approval Links shows "link expired" when email scanners prefetch approval emails @majewskibartosz
  • feat(mcp-client): add piece to call external MCP server tools @sanket-a11y
  • ZenMux: add Management API usage with five-hour and weekly quotas, subscription expiry, and USD PAYG balance. Thanks @kays0x!
  • Settings: add a local Usage & Spend view with honest 7/30-day coverage, native-currency grouping, and exact-home Codex account scans. Thanks @Chipagosfinest!
  • Settings: add a private local share card for Usage & Spend, with native-currency totals and sanitized plan/model labels. Thanks @Chipagosfinest!
  • Unicity AOS Community Edition 2026.1.1.
  • This product release bundles Astrid Runtime v0.10.1. The exact runtime release identity, WIT commit, and compatibility pins are published in runtime-compatibility.toml and in every archive's…
  • The release contains the 18 signed, installable aos-* capsule artifacts selected by Community Edition.
  • Global model presets overhaul — Model presets are now a single global collection. Presets are reusable setups covering main, utility, and embedding models.
  • Default model updated — The global default, Balance preset, and onboarding defaults now use OpenRouter's openai/gpt-5.6-terra.
  • LaTeX rendering in chat — Standard inline and display TeX delimiters are now converted and rendered in chat messages, including agent-thought math.
  • fix(core): ignore CNAME entries in node:dns resolved-address validation
  • fix(server): surface OAuth provider errors on callback
  • feat(github): add 59 actions across repo, issue, PR, CI, release
  • Registry skills content store and manifest for skill discovery
  • Config schema exposed through ACP
  • Checkpointer engine state model, owner union, and pending hunks
Ziva
  • Godot 4.2+
  • Active Ziva account
  • Faster and more reliable ingestion for large files and document batches, reducing wait times when adding data.
  • Improved search relevance that returns more accurate results for typical queries.
  • New dataset-level overview index (optional) that groups documents by topic and creates short summaries so searches have broader context.
  • feat!(openshell-cli): remove openshell policy prove command and z3 de…
  • fix(server): persist sandbox labels on create
  • fix: remove mentions of bundled-z3 in CI and wheel builds
  • CreateSession
  • SendInput
  • FindTurnByClientSubmitID
  • Promote Skills Repository out of experimental status
  • Add Flows in Studio documentation
Strands Agents
  • feat: add community agent governance kit strands invervention handler (4297fb1b)
  • feat(py): add unified storage interface (1aff2707)
  • feat(context-offloader): auto-namespace unified Storage under offloader (d6a2636c)
  • You no longer need to know Cron syntax to create a custom schedule. A visual builder covers interval, daily, weekly, and monthly rules, with a live preview of the generated expression — and the…
  • A new queue-protection toggle keeps scheduled runs from piling up: if a previous run is still going, the next one waits instead of stacking.
  • Every skill now has its own detail page showing its info and which assistants use it — and you can attach or detach assistants right there. The skills list shows used-by avatar stacks at a glance.
  • #1822 a5c568d Thanks @liruifengv! - Add the /copy slash command to copy the last assistant message to the clipboard.
  • #1824 bfecd01 Thanks @liruifengv! - Using an API key for Kimi coding models now also fetches the latest model list automatically.
  • #1811 cec15e2 Thanks @liruifengv! - Fix Esc and Ctrl+C cancelling compaction instead of closing an open /btw panel.
  • A builder for creating Dify Agents
  • You can build Dify Agents through the UI: set a base prompt, upload Skills and files, and connect tools and knowledge from the Dify ecosystem.
  • We also provide an agent that helps you build Dify Agents. Through conversation, it can configure the Linux sandbox environment, install required packages, and create new Skills and files that…
  • Turnkey secure remote / team server. mempalace serve exposes the full MCP surface over HTTP with secure defaults: non-loopback binds require a bearer token, generated tokens are stored with…
  • Milvus storage backend. The new opt-in milvus backend supports Milvus Lite, self-hosted Milvus, and Zilliz Cloud, with COSINE vector search, native BM25 lexical search for new collections…
  • Atomic knowledge-graph fact replacement. supersede() / mempalace_kg_supersede closes an open fact and opens its successor at one shared instant.
  • 统一 CLI、Web 与 Headless 模式的 Agent 内核和事件协议,同时保留原有 Paper2Code 工作流。
  • 新增交互式 CLI、原生编程工具、持久化 Session、项目记忆与长对话上下文管理。
  • 引入权限控制、安全沙箱、工具审批及工作区隔离机制。
  • v0.61.2 → v0.61.8 is all about clean release momentum—promoting the latest work from main into release so staging/prod builds can pick up the newest improvements. Onward and upward. 🚀
  • — Thank you @M3gA-Mind!
  • Thanks to everyone who contributed to this release: @M3gA-Mind, @github-actions[bot].
  • 910671b1853f8cea95c8c5ffdb4889701ab1dbfa Fix Codex task MULTICA_TOKEN passthrough (ZIC-82)
  • ce15300493b2d5c0f597ccc2697d980ed146d2dd MUL-4884: feat(issues): anchor the working chip on agents + real colour tiers
  • ea8ccf31237b15c6cceeed346f5262e5dfbb85bd MUL-4903 fix(agent): harden Cursor terminal failures
  • fix(mcp): migrate ${VAR} headers to env credential refs during driver migration
  • refactor(ci): harden desktop workflows and drop legacy verify dead code
  • fix(memory): restrict automatic memory to external user queries
  • Superserve Tools: Added SuperserveTools to run agent-generated code and manage files in Superserve, a Firecracker-based sandbox platform for long-running agents. See docs.
  • Plivo Tools: Added PlivoTools to send SMS, make voice calls, and look up phone numbers with Plivo. See docs.
  • The Context Company: Added an observability integration to trace agent runs with The Context Company. See docs
Browser Use
  • Release 0.13.6 with Browser Harness 0.1.6
  • Browser Harness can now record browser tasks and turn hours of agent work into short, verifiable videos.
  • The videos use real browser frames, clicks, cursor movement, and typing—then condense the work with concise subtitles, honest mistake-and-correction explanations, and a verified final outcome.
  • Video creation is built directly into Browser Harness. No separate skill or installation is required.
  • ci(release-please): use a PAT so releases trigger the publish workflows
  • ci: bump actions/checkout from 4 to 7
  • fix(memory/sync): make Codex AGENTS.md adapter additive (stop wiping memories)
  • feat: add LiteLLM AI gateway provider preset
  • docs(pages): remove integrations parent page, keep as sidebar group
  • docs(pages): add Delegation Mode documentation to website
  • feat: make task and turn tracing spans configurable
  • feat(realtime): track response usage in session context
  • fix(memory): serialize OpenAI conversation session initialization
  • Add Moonshot AI kimi-k3 model
  • Emit EnqueuedMessagesEvent when enqueued messages are delivered into the run
  • feat: add DeferredToolCallEvent and DeferredToolResultEvent to AgentStreamEvent
  • Fixed argv flag injection via resume and session_id options: --resume and --session-id are now passed as single =-joined argv tokens (e.g.
  • Hardened build scripts against command injection via CLAUDE_CLI_VERSION: Added version validation (_cli_version_validation.py) and eliminated shell interpolation in download_cli.py and…
  • CI now lints and typechecks scripts/ alongside src/ and tests/

Jul 16, 2026

KaneAI
  • OAuth tokens are now exchanged for long-lived credentials automatically — when running tests, OAuth credentials are resolved into non-expiring basic credentials behind the scenes, so sessions…
  • Fallback during credential exchange now warns you — if the auth exchange falls back to a secondary path, you'll see a clear warning instead of it happening silently.
  • Condition blocks now evaluate logical and boolean expressions — the condition leg previously missed logical/boolean expressions; those cases are now handled correctly.
Pi
  • Kimi Coding thinking compatibility — Kimi Coding models now use adaptive thinking correctly; K3 exposes its supported max level and supports replaying empty-signature thinking blocks.
  • Fixed inherited Kimi Coding requests to use Anthropic adaptive thinking effort without token budgets, and enabled empty thinking signatures for K3 and kimi-for-coding.
  • Fixed inherited Kimi K3 pricing metadata for Moonshot AI and Moonshot AI China.
  • Plugins can ship apps. Plugins can now bundle full apps that open right in your library.
  • Your assistant shows off the skills it taught itself. When your assistant writes a new skill through its own memory (the retrospective process that turns patterns into reusable skills), it now…
  • Settings, reorganized. The whole Settings page got rebuilt. General is now split into clean Profile, Version, and Appearance sections with 2FA and a Danger Zone.
Buzz
  • feat(desktop): onboarding step dots, identity/key page restyle, overflow scroll (f2522995c)
  • chore(desktop): add AppShell.tsx file-size override to unblock main CI (db57bc8a2)
  • feat: add invite QR and mobile direct join (648cbf361)
  • Gate artifact URL re-minting to Skyvern signed content URLs (8ed3be1b3)
  • Refuse Copilot drafts whose entire output contract is a null blocker (SKY-12539) (3fcac5d83)
  • chore(deps): bump mcp to 1.28.1 to fix 3 high-severity Dependabot alerts (14439528f)
  • Update GCP Skill Registry to use agentregistry/skill endpoint (fa884e3): Existing users of GCP Skill Registry should note that the underlying endpoint is now AgentRegistry, so users will need to…
  • add Cloud Run sandbox option in code executors (5b1088a)
  • Add search agents and search MCP severs to agent registry (ed579c1)
  • 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)
  • Add organization ID parameter to PlusAPI client
  • Add step interception points and rework execution hooks documentation around @on
  • Wire execution-boundary interception points
  • Widened eve compatibility for @agent-browser/eve to accept eve 0.23 and future major releases without peer-resolution warnings
  • Standardized eve branding to use lowercase styling across the docs, examples, package readmes, and release notes
  • chore: unify secret-redaction key matching in one shared helper
  • chore: update chatgpt login model wording
  • fix: preserve OPENAI_BASE_URL when loading and saving environment
  • [docs]: add docs for setDomainPolicy
  • fix: silence AI SDK "system message in messages" warning in act/extract/observe
  • [STG-2510] fix: source eval browser skill from packages/cli browse SKILL.md
  • Remove redundant subagent depth limiting
  • fix(deepwork): enforce saving deliverables outside internal .slim scratchpad
  • feat(kitty): add kitty terminal multiplexer backend
  • Added an Up Arrow shortcut to close the subagent picker when the first item is selected.
  • Fixed home page scrolling so sticky headers and the session list behave correctly.
  • Fixed startup readiness so WSL server loading is included before the desktop app reports ready.
  • fix(sdk): initialize sandbox annotations
  • fix(driver-vm): run sandbox supervisor as guest pid 1
  • feat(ci): introduce merge queue
No release notes published.
  • Changes since langchain==1.3.13
  • release(langchain): 1.3.14 fix(langchain): only retry retryable exceptions in ToolRetryMiddleware feat(langchain): ToolErrorMiddleware
  • Fixed an issue where submitted messages might not be properly acknowledged, ensuring reliable saving.
  • Resolved blocking behavior during long context processing, improving conversation responsiveness.
  • Optimized session lifecycle management for more stable multi-turn conversations.
  • Reindex 现在会按配置的 VLM 并发数创建 SemanticProcessor,避免维护任务绕过全局并发控制。
  • Reindex 写回语义和向量记录时会根据目标 URI 切换到内容 owner 上下文,避免 admin 或维护任务把用户资源、memory、skill 的 owner 改写掉。该改动同时为 Python、TypeScript、Go 客户端暴露了 dry_run 参数。
  • 新增 prune_orphans reindex 模式,可扫描 resource、memory、skill 向量记录,识别源文件、.abstract.md、.overview.md 或 memory chunk 已不存在的孤儿记录,并支持 dry-run 预览待删除数量。
  • SQLite 本地索引:会话列表、全局搜索、活动统计、定时任务运行记录和 Trace 使用可重建的本地投影加速读取,启动后会持续对齐源文件,并在索引不可用时保留文件回退路径。
  • 完整工作区文件搜索:文件面板可以直接搜索尚未展开或加载的目录,不再需要先手动展开目录树;搜索结果支持继续预览和加入聊天上下文。
  • 内置跨平台 ripgrep:桌面安装包为 macOS、Windows 和 Linux 的 x64 / ARM64 架构准备对应 ripgrep,不再要求用户另行安装 rg 才能使用相关搜索能力。
Kimi Code
  • #1776 ffaf0b9 Thanks @sailist! - Expand the coder subagent tool set to include background tasks, todo lists, plan mode, skill invocation, and nested agents, mirroring the main agent's capabilities.
  • #1771 b513975 Thanks @liruifengv! - Optimize the unit formatting of the context usage display.
  • #1765 d531398 Thanks @RealKai42! - Fix Kimi-provider models routed through the Anthropic protocol incorrectly showing reasoning effort options.
  • feat(fivetran): add Fivetran provider
  • feat(railway): add provider
  • feat(speechmatics): add provider
  • 983545b34d595f33ea762e4d905bd9379d697d24 KAP-867: retry safe Codex initialize timeouts
  • d74aed1e07d32a16ff3be770cd3e4e7f0bcf8a1a MUL-4302 Remove legacy attribution invariant exemption
  • f8654b37c0e0ed20d2b9bb8778306563d4511566 MUL-4817: open issue tabs in dedicated windows
  • fix(gemini): fix gemini schema containing null
  • feat(workspace): add K8sWorkspace with Pod/PVC lifecycle, tar-stream backend and MCP gateway support
  • feat(rag): add MongoDB vector store backend Closes
  • Vulnerability: Fixed a system file permission flaw that allowed anonymous access to resources;
  • Vulnerability: Fixed an IDOR vulnerability in admin chat logs which allowed bypassing application path authentication to read cross-application chat records;
  • Vulnerability: Added independent authorization verification mechanism for agent and workflow tool scheduling;
  • Introduces the command-first zg query, zg index, zg status, zg install, and zg uninstall interface.
  • Adds managed ripgrep compatibility and shared file-selection controls across indexed and no-index search.
  • Preserves index filters across refreshes and improves cross-platform result normalization.
  • [fix] Link local path imports
  • [agent] Capture async generation failures in agent feedback
  • [fix] Centralize editor undo history
  • feat(delegate): add delegation mode for host-agent driven code review
  • feat(rules): add FreeMarker (.ftl/.ftlh/.ftlx) review support
  • feat(llm): add built-in Ollama Cloud provider preset
  • Run health names the problem. Stalled, stale, orphaned, quota-parked,
  • Hijack a live agent session in the browser. Every node exposes its real
  • Transcripts stream live and stay pinned to the newest output while the
  • feat: export HistoryProcessor and add actionable hints to usage-limit and tool-retry errors
  • Fix MistralStreamedResponse streamed number/integer validation to accept int-valued numbers and reject bools
  • fix(onboarding): suggest Agent('test') when provider credentials are missing
  • SOP (Standard Operating Procedure) engine lands end to end — a daemon maintenance tick, typed step contracts, live step execution, cron/filesystem/calendar triggers, and an out-of-band approval plane.
  • WebAssembly plugin host (wasmtime component model) for tools, channels, and memory, with per-call execution limits, signature-policy enforcement, and registry search/install by name.
  • Git forge channel with GitHub and Gitea/Forgejo providers, plus a unified git_forge tool and SOP ingress.
  • Added a /codesearch command (aliases /cs and /search) to interactively search code across your workspace, interpreting queries as regex by default with -F/--literal for exact matching and f:/file…
  • Added copy-on-select in no-flickering mode so dragging highlights text and releasing the mouse copies the ANSI-stripped selection to the clipboard, and hides the virtual scrollbar so it no longer…
  • Added an indicator at each context-compaction boundary so you can see where earlier compaction happened.
  • Fixed: crush auth now works outside of client/server mode
  • New: Crush now auto retries auth under the hood when it fails
  • Improved: dialogues now render great on tiny screens

Jul 15, 2026

  • Updated bundled Claude CLI to version 2.1.211
  • Eve extension - Added @agent-browser/eve, an Eve extension that mounts the agent-browser tool set with namespaced browser tools, sandbox bootstrap helpers, docs, examples, CI, and release packaging
  • Hardened domain allowlists by blocking WebRTC bypasses, applying network containment across launch modes, workers, popups, restored state, and reused daemon sessions, rejecting unsafe startup…
  • Fixed completed-page waits so load and DOMContentLoaded waits resolve immediately when the current document is already ready, with structured Eve wait timeout handling and focused coverage
No release notes published.
  • Agent: Added support for searching (cmd-f on macOS, ctrl-f on Linux/Windows) inside of terminal threads.
  • Added "Agent Panel" and "Git Panel" items to the "View" menu.
  • Improved the agent panel's turn-end actions: thread feedback is now also available via "Positive Feedback" and "Negative Feedback" slash commands (typed as /helpful and /not-helpful) in the…
Buzz
  • Polish agent and team sharing and snapshots (2f3fa586f)
  • fix(desktop): filter punctuation from avatar initials (63e099709)
  • fix(desktop): preserve snapshot attachment URLs (1ab7bbfe4)
  • Added session-modal popup floating terminal panes for type = "popup" custom command keybindings and plugin panes, with optional cell or percentage sizing and no changes to the tiled tab layout.
  • Added ui.copy_on_select to disable automatic clipboard copying after mouse selection while keeping the selection visible.
  • Added configurable row layouts for expanded Space and Agent sidebar entries, including built-in display tokens, per-agent overrides, custom metadata tokens, and pane/workspace metadata reporting…
  • Stopped subagents from launching nested subagents by default, with a configurable subagent_depth limit when needed.
  • Improved default reasoning depth for Meta models.
  • Added Mod+N as another shortcut for opening a new tab.
  • fix(frontend): restore cross-domain PostHog tracking by aligning client/server distinct_id (WIP)
  • fix(app-server): lower DB pool defaults and make them env-tunable
  • fix(mcp): preserve MCP auth secrets stripped by settings GET round-trip
bb
  • BB Official plugin release for bb-plugin-memory@0.2.0.
  • None noted in this changelog.
  • Added session scoping to AgentController so independent sessions can run in parallel over the same resource (for example one session per git worktree).
  • Added support for custom processors on createScorer judges. You can now pass inputProcessors, outputProcessors, errorProcessors, and maxProcessorRetries in a scorer's judge config…
  • Keep reminders out of cacheable system prompts
  • @DanMaly made their first contribution in
  • Warning when the Narrator is enabled without working audio output
  • Windows command prompt is now shell-aware
  • Hardened bash tool permissions and cross-platform shell handling
  • 轻量记忆模式配置:支持通过配置控制记忆演化流程,便于在轻量场景下降低后台处理开销。
  • L3 抽象模型配置:新增专用的 L3 LLM 配置入口,用于独立管理抽象总结阶段的模型调用。
  • 中文关键词召回支持:增强 CJK 关键词分词能力,提升中文内容的检索与召回效果。
  • 5b26b99722d312a0bade380acf4f2d1229de271f MUL-4164: support Intel macOS Desktop packages
  • 54fd29ebddfd80c2d1531006aaf71b57d4a048d4 MUL-4383 fix(daemon): stop routing CodeBuddy skills/memory through Claude's .claude paths
  • 51e4afc4800e978789bf30d57dc08457afa80df6 MUL-4753 Fix Lark recent context fetch degradation
  • #1704 38a2363 Thanks @sailist! - Align the print-mode run lifecycle across engines: print_background_mode and print_max_turns now take effect for kimi -p on the experimental engine, with the same…
  • #1704 38a2363 Thanks @sailist! - Align the subagent timeout across engines: a fixed 2-hour default, overridable with [subagent] timeout_ms in config.toml or the KIMI_SUBAGENT_TIMEOUT_MS…
  • #1727 286d3e7 Thanks @liruifengv! - Add a builtin check-kimi-code-docs skill that answers Kimi Code product questions (CLI usage, configuration, membership, error codes) against the official…
  • No commits or diff statistics provided; no user-visible changes.
  • refactor: preserve app reference source order
  • macOS Apple Silicon (arm64)
  • Thank you @nazarli-shabnam for improving resilience by making /announcements/latest return null on 404 instead of erroring.
  • PyPI: bash pip install crawl4ai==0.9.2
  • Docker: bash docker pull unclecode/crawl4ai:0.9.2 docker pull unclecode/crawl4ai:latest
  • Note: Docker images are being built and will be available shortly. Check the Docker Release workflow for build status.
  • 0.6.9 changelog
  • [feat] add rich text styling to agent tools
  • fix agent clip placement undo boundary
  • feat: Add OpenAI Responses API support and refactor protocol handling
  • Make message_history provider-valid out of the box (repair tool-call/result pairing)
  • Add support for OpenAI background mode and handle Anthropic stop_reason=pause_turn
  • Fix has_values() returning True for all-zero usage details

Jul 14, 2026

Claude Agent SDK
  • Updated bundled Claude CLI to version 2.1.210
  • Vellum can handle your speech now. Speech-to-text and text-to-speech used to mean going and getting your own API keys from a provider like ElevenLabs or Deepgram.
  • GPT-5.6 with prompt caching. GPT-5.6 models are now available. They support prompt caching, which means the big chunks of text that get sent every turn…
  • Touch gestures on mobile. Swipe left on a conversation to reveal delete, pin, or rename. Press and hold a message to get copy, edit, or retry.
Buzz
  • fix(desktop): allow staged updater commands (109c2c526)
  • feat(teams): unify team model, snapshot sharing, and PNG memory parity (448baeef7)
  • fix(desktop): always start agents and hide unavailable mesh (37bc962cb)
Open Interpreter
No release notes published.
OpenCode
  • Fixed spacing between model provider sections in Settings.
  • Godot 4.2+
  • Active Ziva account
Palmier Pro
  • Add v0.6.7 to appcast
bb
  • bb desktop 0.0.30
  • fix(tui): redraw after sandbox shell exits
  • fix(agents): add confirmation gate to triage-issue batch mode
  • fix(gator): retry review after draft blocker clears
  • Removed the openai-responses compat.sendSessionIdHeader flag from models.json. Session-affinity behavior is now controlled by compat.sessionAffinityFormat…
  • Cache-friendly dynamic tool loading - Extensions can add tools during execution while supported Anthropic and OpenAI Responses models preserve prompt-cache prefixes. See Dynamic Tool Loading.
  • Message copy shortcut - Ctrl+X copies the last assistant message in the transcript or the selected message in /tree, making older and branched messages directly copyable.
  • Observability support for Agent Teams – Added full‑chain observability for Agent Teams with OTEL trace export
  • Team permission control – Implemented team‑level permission management
  • Worktree isolation for teammates – Aligned teammate worktree isolation with the AgentTool flow
  • Added ValkeyDb as a fast in-memory database for agents, teams, and workflows. See Docs
  • Added a ValkeyDB vector store with vector and keyword search. See Docs
  • Redmine Tools: Added RedmineTools to manage issues, comments, and time logs in Redmine, an open source project management tool.
  • Codex Windows hooks kept failing after v3.4.1 and v3.5.0: root cause found. The shell resolver preferred C:\Windows\System32\bash.exe, the WSL launcher, whenever Git Bash usr\bin was not on PATH…
  • Windows Python discovery hardened. pwf-hook.cmd now honors PYTHON_BIN, probes standard uv and CPython install locations when Codex starts hooks with a reduced PATH, and quotes the interpreter path…
  • Pi extension 1.2.1: recitations no longer break interactive tools. Messages queued by the tool_call hook with deliverAs: "steer" were consumed as the answer of an open AskUserQuestion dialog.
  • feat: more sensitive files & allow read global
  • chore: bump version to 2.0.0post2
  • test(unit): runtime/security/install regression tests — PR-A4
  • Gateway local L2 定时器未走场景提取:本地 timer 扫描将 L2 schedule 错路由到非 scene extraction 任务类型,导致 L2 场景提取定时触发失效。抽出 timer-routing.ts,按 timer member 前缀正确映射 offload-l1 / offload-l15 / offload-l2 与 L1 / L2 / L3 任务。
  • 升级 cos-nodejs-sdk-v5 至 3.0.0:旧版 COS SDK 存在安全风险,升级从而修复该问题。
  • Sending to a busy conversation is handled gracefully across all runtimes: instead of surfacing a raw conflict error, AionUi now recognizes "conversation is busy" responses and recovers the queued…
  • Fixed messages being truncated when the text happened to mention the internal file-attachment marker — attachment parsing now only applies to actual attachment blocks on your own messages.
  • If assistant data fails to load at startup, AionUi now treats it as a repairable local-data issue and guides you through recovery instead of failing with a generic error.
  • track Agent GUI engagement.
  • backport Claude binary runtime provisioning.
  • [CP] feat(agent): track Agent GUI engagement
  • sub2api: add group-key usage with daily, weekly, and monthly quotas, multi-account switching, wallet balance, and expiry details. Thanks @weirdo-adam!
  • Kimi: reuse fresh signed-in Kimi Code CLI credentials in Auto mode without refreshing or rewriting CLI-owned authentication state. Thanks @Leechael!
  • Community integrations: list codexbar-plasmoid, a KDE Plasma 6 usage widget. Thanks @psimaker!
  • fix(app-server): preserve conversation created_at across lifecycle webhooks
  • fix(mcp): preserve SaaS credentials with encrypted storage
  • ci: wait for the docker build before retagging images
  • Weak-password detection skill
  • LLM prompt-injection skill
  • GCP and Auth0 reconnaissance skills
  • e1d0d68c53db0232993a504d1983e191eb6c472f Fix public host root redirect
  • ef9b334408789964e3f3be3746433905aaa27f56 MUL-4398: fix Hermes bound-skill discovery with per-task overlay
  • 5b57a8ebcae6decda639d75f91c7eb6040710eb4 MUL-4460: fix(email): add SMTP_FROM_EMAIL for SMTP sender
  • Reconnect desktop ACP sessions after sleep and connection loss
  • Per-message usage stats UI (tokens, cost, TTFT, tok/s)
  • Per-message usage/cost tracking with derived session totals
Kimi Code
  • #1678 ec1c974 Thanks @chengluyu! - Preserve goal completion summaries and show untyped LLM errors without an internal error-code prefix in step interruption events.
  • #1688 94c0ef8 Thanks @sailist! - Fix built-in tools being unavailable when the model provider becomes ready after the session starts.
  • #1684 e417ee7 Thanks @RealKai42! - Fix Kimi sessions getting stuck when preserved-thinking history contains an empty reasoning step.
  • npm: bash npm install -g base44@latest
  • Homebrew: bash brew install base44/tap/base44
  • Version: 0.1.4 NPM Tag: latest
  • Grok 官方登录:新增 xAI Grok 官方 OAuth 登录、token 刷新、动态模型目录和会话运行支持,内置 Grok 4.5 与 Composer 2.5 fallback,可直接在桌面端完成登录、选模和退出。
  • Claude Code Auto 模式:权限选择器新增 Auto 模式,启用前会明确提示风险;运行时会审查工具调用和提示词注入,自动执行其认为安全的操作并阻止高风险操作。
  • 内联 Diff 评审:工作台新增完整 Diff 审阅界面,支持语法高亮、旧/新行定位、单行或连续行评论,并把评论和代码引用带回聊天输入框。
  • fix(xlsx) Caching of formats, fonts, fills and borders on Excel rendering
  • 修正 DOCX skill 的分页和 TOC 指引
  • @andulv made their first contribution in
  • feat(telemetry): propagate W3C traceparent from parent process
  • fix: remove redundant font-family from .response-text .inline-code
  • Fix typed composition members recursion in JsonSchemaTransformer
  • fix: handle int timeout values in MistralModel._get_timeout_ms
  • Remove retired models from KnownModelName and restore Gateway-supported models

Jul 13, 2026

  • Added an f (full diff) shortcut to the create-file tool review screen so new-file confirmations can open a full-screen diff view, matching the existing file-edit experience.
  • Added support for pasting the OAuth authorization code in print mode (-p) via the controlling terminal (/dev/tty on POSIX and CONIN$ on Windows) when stdin is consumed by a piped prompt, and made…
  • Improved responsiveness on large conversations (5000+ steps) in no flickering mode by switching hot-path line-count methods to pointer receivers, cutting the per-frame prefix-sum cost and…
  • Control UI overhaul: conversations are easier to organize and work with side by side, with live Tasks, clearer chat controls, better usage and cost views, files, downloads, pairing, approvals, and…
  • Easier setup from install to first chat: guided setup provides clearer steps toward a working first chat, checks connections before saving them, and preserves earlier choices when setup is…
  • Official app updates: iOS and iPadOS, Android, and macOS received substantial work across setup, navigation, chat, voice, permissions, localization, files, scheduled work, offline reading, queued…
Buzz
  • fix(desktop): unify observer feed scroll onto useAnchoredScroll (d75c2e913)
  • fix(desktop): sync sidebar update card copy with global installing state (b63a2e423)
  • fix(desktop): parse codex ACP plan entries[] into checklist (34dcd13d5)
OpenCode
  • Removed an obsolete Codex workaround that could interfere with OpenAI Luna Responses Lite requests.
  • Updated Azure AI support for GPT-5.6.
  • docs: update OpenWiki
  • chore: verify node.js 24 compatibility in ci
  • feat: add nvidia nim as a model provider for inference
  • WebGPU launch preset - Added --webgpu across the CLI, config, environment, and MCP surfaces, with hardware backends on macOS and Windows, software Vulkan on Linux, automatic Xvfb for displayless…
  • Added periodic restore-state autosaves while the browser remains open, preserving recent state after a browser window is closed by hand and capturing background page changes while honoring the…
  • OllamaCloudProvider with dynamic model discovery and context limits
  • Per-message usage stats UI (tokens, cost, TTFT, tok/s)
  • Per-message usage/cost tracking with derived session totals
  • #2283 871ca7e Thanks @seanmcguire12! - add context.setDomainPolicy({ allowedDomains: ["allowed.domain"] }) which allows users to define a set of domains that are accessible to stagehand
  • #2274 f31980f Thanks @seanmcguire12! - add context.setDomainPolicy({blockedDomains: ["some.domain"]}) which allows users to define a list of domains that will be blocked by stagehand
  • #2305 cd1daad Thanks @shrey150! - Remove the noisy AI SDK "system message in messages" warning logged on every hybrid/DOM agent.execute() call.
  • #1902 a9d78c0 Thanks @mattzcarey! - Always apply the Worker-safe CfWorkerJsonSchemaValidator to MCP client connections by default.
  • #1903 3ba6a78 Thanks @mattzcarey! - MCP client: url-mode elicitation support with a real elicitation handler
  • Agents can now respond to server-initiated elicitation/create requests by
  • docs: update README for OpenAI GPT-5.6 spotlight
  • refactor: remove dead code and unused exports
  • refactor: extract duplicated code into shared utilities
Mem0
  • Platform backend: Encode dynamic URL path segments so memory and entity IDs containing special characters no longer produce malformed requests
  • fix(tui): route warning logs to status bar instead of stderr
  • fix(mcp): include tool names in policy logs
  • chore(deps): bump astral-sh/setup-uv from 8.3.1 to 8.3.2
  • Fixed "409 conflict" errors when sending queued messages: the draft box could fire a message while the backend was still wrapping up the previous turn…
  • Team runs no longer flash a misleading "queued" state right after you start them — accepted work shows as processing immediately, while genuinely queued items still display their queue position.
  • fix(conversation): prevent queue drain from racing backend idle state
  • chore: bump version to 2.0.0.post1
  • fix(models): prevent browser autofill on provider search input
  • fix: fix legacy session loading for file content blocks
  • 12e3c393d7e51df3040805b95c80ddbdcbcd0a73 Add auto-save confirmation toasts
  • 41b3045efa7599187f944b67cbd0eab438542a15 MUL-4424: bound Codex app-server startup RPCs
  • 3bd69bdaa2ec64d0a26db4308acf011f6eafc835 MUL-4432: add DM button on agent detail page
  • feat(pages): add blog section with i18n support
  • fix(llm): support opt-in OpenAI streaming
  • fix(background-file) Manage path at git repository root directory Refs:#324
OpenViking
  • 新增 Cursor 集成:安装器一次配置生命周期 Hooks、MCP、Rule 和 memory Skill,实现会话启动与请求前召回、响应后捕获。
  • 新增 TRAE 与 TRAE CN 集成:支持 SessionStart、UserPromptSubmit、PreToolUse 和 Stop Hooks,在短会话结束时也会立即提交记忆。
  • OpenCode 插件升级为 MCP 与生命周期 Hooks 的混合模式,支持仓库上下文、自动召回、会话捕获和共享凭证配置。
OpenSandbox
  • Align with sandbox SDK 1.0.16 — Bumped the Kotlin code-interpreter SDK artifact to 1.0.16 and bumped its com.alibaba.opensandbox:sandbox / sandbox-api dependency to 1.0.16.
  • Fix Maven POM repository URL — SCM / project URL in the published POM now points at opensandbox-group/OpenSandbox instead of the old alibaba/OpenSandbox.
  • @Gujiassh
  • feat(provider): add PubMed
  • feat(provider): port providers
  • feat: upgrade typescript to version 7.0.2
  • Switch holmes image to Alpine base
  • Bound conversation/tool-call claiming to worker capacity (ROB-388)
  • ROB-306: github-mcp PAT path uses native HTTP mode
  • npm: bash npm install -g base44@latest
  • Homebrew: bash brew install base44/tap/base44
  • Version: 0.1.3 NPM Tag: latest
  • Added support for importing Claude-exported conversations to facilitate data migration.
  • Introduced an external session import wizard to guide users through importing non-resumable sessions.
  • Enhanced banner notifications for unavailable sessions with clearer and friendlier messaging.
  • Codex hooks on Windows emitted invalid JSON and failed on Unicode (PR
  • Closed and complete Pi plans kept nagging "Task incomplete"
  • Four language commands referenced a skill namespace that does not exist; corrected to planning-with-files:planning-with-files-<lang>.
  • wren: serve project capabilities over MCP (a38bfd7)
  • clickhouse: sanitize brackets in connection-URL userinfo before parse (ef0a824)
  • clickhouse: use unquote (not unquote_plus) for URL credentials; decode database (d9ba1a1)
  • Fixed an issue with npm v12 that could cause language server startup failures

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.