AI agent releases: Jun 29 – Jul 5, 2026
137 stable releases from 71 tracked agents and frameworks in the week of Jun 29, 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 5, 2026
- Claude: show opt-in read-only claude-swap accounts as stacked usage cards without delaying ambient refreshes. Thanks @optimiz-r!
- Claude: switch inactive claude-swap accounts directly from their stacked usage cards and refresh usage immediately.
- Codex dashboard: show calendar-correct raw Today and 30-day credit totals without converting credits to billed dollars. Thanks @avenoxai!
- feat: init openwiki docs
- ci: Add openwiki gh action
- fix: improve streaming
- Alpha Vantage look-ahead filter now runs. The fundamentals payload is a JSON string, so the dict-only guard silently skipped filtering and future-dated reports leaked into historical runs.
- News analyst prompt matches the tool. It advertised get_news(query, ...) while the tool takes a ticker, so the model hallucinated free-text query calls. (#1116, @shcheuk)
- Shared debate/risk routers can't crash mid-run. Both routers returned more targets than any one edge mapped; every edge now shares the complete path map. (#1088, @Fr3ya, @sa7an7, @Sushanth012)
- Fall back to stopped snapshot containers — When committing a Kubernetes sandbox snapshot, the image-committer now falls back to nerdctl ps -a if the running-container lookup returns empty.
- @Gujiassh
- Docker Hub: opensandbox/image-committer:v0.1.1
- fix: forward DeepSeek V4+ reasoning_effort for openai-compatible providers
- fix(model): support Claude Fable series and Opus 5 in capability detection
- fix(models): add Anthropic native /v1/models fetcher(target V1 branch)
- 📄 Repo context in prompts — PR-Agent now feeds AGENTS.md to /review, /describe, and /improve by default, so the AI picks up your project conventions out of the box.
- 🧩 Agent Skills (SKILL.md) — package reusable, domain-specific guidance and expose it to the AI by pointing [skills] paths at it.
- 🏢 Organization-level settings — manage configuration for a whole org from a single pr-agent-settings repo, on GitHub, GitLab, and Bitbucket.
Jul 4, 2026
- Codex: show every available reset-credit expiry in menus and provider settings, including non-expiring credits, and summarize credits nearing expiry. Thanks @brahimhamichan!
- Cost history: optionally show shorter 7, 30, and 90-day comparisons from the selected local history window. Thanks @jtl06!
- Codex cost history: group local usage and costs by project and worktree in menus and CLI output. Thanks @clemenspeters!
- feat(connectors): add Snowflake target
- feat: support multiple GPUs and fractional GPU allocations via GPUPool
- feat: add type_hint to use_state for typed state deserialization
- 0.6.1 changelog
- feat(timeline): multiple timelines per project
- Audio Enhancer / Denoise
- Added TogetherSTTService and TogetherTTSService for real-time speech-to-text and text-to-speech using Together AI's WebSocket APIs.
- Added per-sentence synthesis mode and zero-shot audio prompt support to NvidiaTTSService, letting NVIDIA TTS users choose between stitched and per-request synthesis flows and configure…
- Added on_heartbeat_timeout event handler to PipelineWorker, fired when a heartbeat frame is not received within the monitor timeout period.
- The bespoke gateway-WS collection stack is retired. gateway-client / gateway-react now sit on TanStack DB collections via createSmithersCollections with per-mode providers.
- AI SDK 7 upgrade. onStepFinish is renamed to onStepEnd, and Smithers now requires Node >= 22.
- Manage it explicitly with bunx smithers-orchestrator gateway status and bunx smithers-orchestrator gateway stop.
- Fixed a race condition in nested agent auto-completion, resolving input lag or response anomalies.
- Scoped single-provider rail sections to provider-local targets, preventing cross-module interference.
- Enhanced the reliability and smoothness of agent conversations.
- #1367 23daf0f - Revert the recent TUI transcript rendering changes to the original upstream behavior while rendering issues are reworked.
- #1343 ec758c7 - Add click-to-enlarge for images uploaded in the web chat. Click an image in a message to open it.
- #1343 ec758c7 - Fix uploaded videos failing to play in the web chat.
- fix(fallback): resolve three deadlock points in foreground fallback + task-session-manager
- feat: compact sidebar mode for agents panel
- Fix dashboard shutdown resource cleanup
- 1ae0a1f5bf545740a86bb7536879989eef3f375a fix(agents): defend AccessPicker against undefined invocation_targets (GH #4915)
- eea65118c15b9e9a2f917c2346b1152deea27f1d feat(gate): turn the smart remote-migrate gate on by default
- 75eafb6d4394eaddaaef75a3ce1b52a4a9da78ea feat(prime): remind when AGENTS.md and CLAUDE.md diverge
- 527ea0c23bca76c0270eee29738320999e1ddf01 fix(embedded): let working-set reconcile commands open past dirty-table migration guard
Jul 3, 2026
- autogen.beta promoted to the top level ag2.x – The framework built up under autogen.beta across v0.12 → v0.14 is now the mainline framework in this repo.
- Classic framework removed – The classic agents, orchestration surfaces, and interoperability layers that were deprecated in v0.12.0 and removed from this repo in v0.14.0 are now fully absent.
- Classic CLI removed – The classic command-line entry point is gone with the classic framework.
- feat(agent-savings): land coding + general workload personas on main
- feat(content-router): lossless-excluded compaction (grep/log/json) + enable in coding/general personas
- feat(anthropic): add Claude 5 family pricing & align current rates
- Godot 4.2+
- Active Ziva account
- .NET: Harden dotnet-format workflow shell handling
- .NET: Add AgentSkillsSourceContext to AgentSkillsSource.GetSkillsAsync
- .NET: Foundry Hosting per-user session isolation and Responses v2 protocol fast-fail
- fix(desktop): bound read-state localStorage growth and recover from quota errors (a3cf7eec1)
- Customize macOS DMG installer (bbf8d8912)
- mobile: thread scroll-to-bottom and desktop-parity mention autocomplete (e9318f66d)
- #1353 68ad686 - Fix duplicated transcript content appearing in scrollback during streaming.
- #1340 e2fe62a - Fix sessions silently dropping later user messages after a turn was interrupted between a tool call and its result.
- #1342 84d8d5b - Have context-compaction notes capture a forward plan for the remaining work — upcoming steps, settled decisions, and foreseeable obstacles — instead of only the immediate next…
- fix(agent): set target id on imported sessions
- macOS Apple Silicon (arm64)
- macOS Intel (x64)
- TwelveLabs Tools: Added TwelveLabsTools to analyze videos and generate multimodal text embeddings. See docs.
- Sofya Tools: Added SofyaTools for search, extract, and research.
- SearchAPI Tool: Added SearchApiTools with Google, News, Images, and YouTube search.
- feat(docker,podman): add SELinux label support for bind mounts
- test(e2e): run rootless podman on ubuntu host
- feat(policy): accept numeric UIDs for sandbox process identity
- session-catchup.py, the mechanism behind resuming a session after /clear, did nothing on Windows.
- inject-plan.sh's containment guard silently dropped plan re-injection and tamper detection on Windows paths reached through an 8.3 short name or the MSYS /tmp mount alias, again with no visible…
- task_plan.md files without ### Phase headings were falsely reported as "0/0 phases complete" (#191, reported
- "Element not found" is no longer a silent failure — when a presence check is uncertain, kane-cli now escalates to a visual scan instead of confidently returning false, reducing missed detections.
- Hover actions work on vision coordinates — previously, hovering over a vision-identified coordinate would fail; it now maps correctly to a click-compatible action.
- Special characters in typed text are handled correctly — literal tokens inside type and fill actions are now properly re-escaped, preventing misinterpretation of characters that would otherwise be…
- worktree NAME option to create or reuse a git worktree and run inside it
- Rewind support over ACP for both the agent and the host
- Teleport URL in the CLI is now clickable
- Full Changelog
- feat: add Astro-specific review rules
- feat(viewer): make human readable token usage for session
- feat(pages): add Docs page with search, markdown rendering, and i18n support
- Doubao: add signed Volcengine AK/SK support for Coding Plan session, weekly, and monthly usage. Thanks @LeoLin990405!
- CrossModel: add API-key wallet balance and UTC daily, weekly, and monthly spend tracking. Thanks @hujuncheng!
- Localization: complete Traditional Chinese provider and menu coverage, and route remaining provider UI copy through localized formatters. Thanks @jack24254029!
- Headline savings is now a flat 65% everywhere. We retired the ~50-65% range and the ~75% claim that still lived on the docs site.
- /caveman-stats stops mislabeling savings (#601, and the budget-% fix). Session token stats are now attributed to the mode that was active when each message happened, so a mid-session /caveman…
- New limit-headroom meter in /caveman-stats for subscription users — session budget saved at a glance.
- 1ff99e5afc1f44ec58a4d3dff3b216a0e7898c07 fix: honor completed codex turns during process eof races
- AutoPause mechanism: A sweeper in the new cube-proxy-sidecar (under CubeProxy/sidecar/) tracks sandbox activity via last_active timestamps reported by CubeProxy's log_phase.lua.
- AutoResume mechanism: When a dataplane request arrives for a paused sandbox, CubeProxy's sandbox_state.lua gate intercepts it and fires an internal sub-request to the sidecar's /internal/resume.
- Configurable resource release ratio: A node-level configuration host.quota.paused_resource_release_ratio (float [0, 1], default 0) controls how much CPU/memory quota paused sandboxes release back…
- fix: repair pref bug for empty preference filed
- Fix #2008: fix: ci test failed
- fix: Tighten preference memory prompt to avoid over-extractions.
- e3464f4388d963e27c702479fb6f3184cb8cea66 Merge pull request #2917 from PierreLeGuen/nearai-provider
- a75b3d15bbadc0850a666cf4beaa1894e2be7925 Merge pull request #3053 from chengzhichao-xydt/codex/store-lock-type-assert
- ec9c6cde8eb67ab94866b0354707ddd40345773a Merge pull request #3091 from chengzhichao-xydt/codex/provider-native-search-type-assert
- Knowledge Base: Added visual progress display for file uploads; files that failed to upload support one-click re-upload.
- Agent: Supported batch configuration for agent conversation log cleanup strategies.
- Agent (X-Pack): Enabled voice input capability for WeCom robot integration.
No release notes published.
- Added complete Spanish (es-ES) and Persian (fa-IR) translations across the whole app, and filled in gaps in the existing locales.
- Skills loaded on a conversation now appear in the / command menu, not just the + menu.
- When a conversation becomes active (via search, a notification, or a link), the sidebar now expands the section and project folder it lives in and scrolls it into view, instead of silently missing…
- 🤹 Polly multi-agent orchestration: the polly example orchestrator now fans work out across more coding sub-agents — cursor, hermes, and opencode join the roster — so a single run can delegate to…
- 🔌 Harness plugin SDK: bring your own coding agent — any harness can now ship as an installable Python package (via the omnigent.community.harness entry point), no fork or core changes required.
- 🧠 Intelligent model routing, now automatic: the model router (introduced in v0.3.0) now picks the best harness and model for each turn via a server-side judge, reading a live per-runner model…
- iFlytek Spark and Astron MaaS providers
- edit session flag to edit conversation before forking
- Fireworks AI declarative provider
Jul 2, 2026
- New onboarding. No more blank slate. First-time setup is now a guided walk-through: name your assistant, pick a personality, connect your tools, and you're ready before your first conversation.
- More distinct personality. Your assistant's personality now comes through clearly and consistently. Same assistant, same vibe, every conversation.
- More succinct responses. Responses are shorter and get to the point faster. If you ask "what's 2+2," you get "4," not a paragraph about the history of arithmetic. Less scrolling, more doing.
- Provider/proxy 连接更稳:provider 请求默认使用 direct proxy mode,只有显式选择 System proxy 时才继承系统代理;第三方 API 返回的 model 与请求模型 ID 不一致时,不再把上下文窗口误显示为 200000;reasoning effort 会正确透传到 proxy transform。
- 权限和首轮会话行为更准确:新增默认会话权限模式设置,bypassPermissions 会跳过可跳过的审批而保留显式 deny 和必须交互的工具;干净安装后添加 provider 的第一轮提问会先应用 active provider runtime,再发起 prewarm 和 user message。
- 历史会话和列表性能继续收敛:历史会话排序不再被 metadata-only 写入刷新成“刚刚更新”;会话时间展示统一为更稳定的月日时分格式;非 Git 项目不会因为新会话 git-info 探测拖慢;最近项目和会话列表加载路径继续减少重复扫描。
- bb desktop 0.0.27
- .NET: Fix typo in XML doc comment for workflow outputs param
- .NET: Add BackgroundTaskCompletionLoopEvaluator for harness background agents
- .NET: Foundry hosted-agent toolbox OAuth consent support
- fix: remove startup-banner emoji that crashes import on non-UTF stdout
- fix(benchmarks): don't mutate shared config across worker threads in get_sb_environment
- @EnesYilmazcode made their first contribution in
- Introduced a deferred first-send model gate: missing-model setup now happens in-thread instead of on the welcome screen, preserving the pending prompt across refresh and auto-dispatching once a…
- Unified the onboarding provider selection into a single step covering Cloud, Account, and Local choices, replacing the previous separate illustrated paths.
- Utility model selection is now explicit — no longer silently synced to the main model — and defaults to each provider's utility model when available.
- feat(helm): add TLS termination for Envoy Gateway ingress
- feat(agents): add manifest-driven gator agent
- #1304 0fc0ae3 - When large images are compressed, tell the model the original and delivered image details.
- #1315 b40bb71 - Fix TUI rendering bugs that caused the screen to go blank and the input box to disappear.
- #1303 2639786 - Fix the TUI crashing when the terminal is resized to a very narrow width while the input contains CJK or emoji text.
- Data source connector: Introduces Google BigQuery as a data source connector for document ingestion and incremental syncing. #15871
- MCP: Adds two MCP tools ragflow_list_datasets and ragflow_list_chats to the RAGFlow MCP server. #15384
- File ingestion
- LocalFileSystemTools
- Read files via the enable_read_file flag.
- Keep file operations within target_directory by default; set restrict_to_base_dir=False to opt out.
- MCP 与本地文件集成增强:MCP 新增 compact description mode,降低工具描述带来的上下文 token 开销;本地文件上传支持通过签名 temp_upload token 自动入库,让 MCP 客户端处理本地文件更顺滑。
- Agent / Bot / OpenClaw 生态改进:Codex 插件增强 upstream marketplace 安装与 recall 稳定性;OpenClaw tool parts 现在会以 assistant messages 形式保存;Bot 增加超时配置,并修复飞书主动消息 metadata、工具调用顺序、thinking mode 和图片上传等问题。
- 存储、加密写入与 QueueFS 稳定性:加密写入改为 temp-file publish + 双路径锁保护,并进一步优化 temp-write 元数据开销;QueueFS 修复 sidecar、图片 rewrite 等路径的锁复用问题,挂载时跳过控制插件缓存;delete/write-protected root 的移动和删除保护语义进一步对齐。
- Full Changelog
- 8b9caded75b6aa23e55a1a5854c0e4f15c11a2aa Fix Mermaid syntax error rendering (MUL-3900)
- 5ed381a9d63ab56e38ddc558f187c19a9014d032 Fix comment attachment URL resolution
- 8b529efc097f37577fe08b87e41384de9947680b Fix recovered Antigravity transcript messages
- fix(fcpxml): flat asset-clips, per-app value encoding, crop/position calibration
- Limit transcription audio extraction concurrency
- Incentivize login with free credits in agent chat panel
- Language selection in desktop app
- French, German, Italian, Portuguese, Indonesian, Malay, Vietnamese, Traditional Chinese desktop locales
- Korean localization for desktop app
- The update dialog detects a codeburn CLI too old to install menubar updates (pre-0.9.9) and shows the exact CLI upgrade command first.
- Local/Combined usage toggle backed by combined multi-device data.
Jul 1, 2026
- Mixture-of-Agents is now a first-class model you can pick — MoA used to be a mode you toggled.
- See every model's reasoning, then watch the answer stream in — When a MoA ensemble runs, each reference model's full output now renders as its own labelled block — you can read what GPT-5 thought…
- The agent verifies its own work — "done" means proven, not claimed — Hermes now records verification evidence for coding work and can decide it's finished by actually running your project's…
- Group consecutive desktop messages (3d08c3b0)
- Update Buzz app icon (db2a9701)
- feat(desktop): add workspace rail unread observer (ce1f13e8)
- Whole-line content now shown in the edit diff
- Declined or skipped tool calls now render as a muted square
- Auto-approve now works in lean mode
- bb desktop 0.0.26
- Added in-thread search to the Agent Panel (Ctrl/Cmd+F). (#57231; thanks dandv)
- Added an entry to the Agent Panel options menu to quickly add a remote MCP server.
- Improved saved thread model selection for cloud-based providers.
- Guaranteed retrieval: Memories relevant to each prompt are now prefetched and injected as a <mem0-relevant-memories> block before the agent runs, so context no longer depends on the agent…
- Multi-hop retrieval: The memory policy and the mem0_memory tool description now push proactive, multi-hop search -- several queries with different phrasings instead of stopping after one.
- Clearer tool schema: The mem0_memory action and parameter descriptions (action, query, content, memory_id, scope) now explain each action and scope.
- @mastra/vercel: sandbox exports renamed—VercelSandbox now means MicroVM-backed Vercel Sandbox.
- Renamed the AgentController interval API. heartbeatHandlers is now intervalHandlers, the HeartbeatHandler type is now IntervalHandler, and the removeHeartbeat()/stopHeartbeats() methods are now…
- Added optional maxSteps to ScorerJudgeConfig and GoalConfig, letting callers control the internal judge agent's agentic-loop iteration limit instead of relying on the implicit default of 5.
- fix(helm): generate namespace-aware SANs in certgen and cert-manager templates
- refactor(core): remove unused extra bind addresses
- docs(mcp): fix granular policy lifecycle examples
- Force reasoning mode for OpenAI-compatible reasoning models so reasoning settings apply reliably on custom deployments
- Stop replaying stale GitHub Copilot response item IDs, avoiding follow-up request failures
- Let question prompts be minimized and restored without dismissing them (@arvsrn)
- Launch Browser Use CLI 3.0 powered by Browser Harness.
- Add browser-use skill so coding agents can install the Browser Use skill directly.
- Pin Browser Harness to 0.1.4 and ship the Browser Use skill with the package.
No release notes published.
- Rename the packaged skill identity to Browser Use.
- Add Browser Use v4 guidance to the skill.
- Make the API key stdin example copy-pasteable.
- Network SSE is now a toggleable mode — a new network_sse flag (and matching TUI toggle) lets you stream server-sent events through the run pipeline rather than waiting for full responses.
- SSE activity is visible in the run view — connection summaries and an analyzer log surface what SSE connections are active, so you can see streaming traffic at a glance without leaving the terminal.
- Replay arms and runs automatically with SSE — when SSE mode is on, replay start and stop are handled for you; no manual setup required.
- 后台任务和 Goal 续跑更可靠:后台任务在前台 turn 完成后继续显示,并进入更清晰的 background task drawer;Goal 自动续跑会插入可见分隔标记,避免多个 assistant turn 混在一起;task-notification-only follow-up 不再污染桌面 transcript。
- 启动、预热和会话刷新更稳:修复 prewarm_session 与 user_message 几乎同时到达时 idle timer 杀掉活跃 CLI 的竞争;查看历史会话不会触发预热;Electron sidecar 会等 /health 真正可用后再继续;session list summary 加缓存,降低大历史列表刷新超时概率。
- Provider/proxy 链路更兼容:CLI-only OpenAI-compatible provider 会启动轻量 standalone provider proxy,不再依赖固定 localhost server;loopback provider target 会绕过系统代理和 Bun proxy fetch options;provider connectivity…
No release notes published.
- docs: changelog and blog post for v0.8.3
- test(retain): serialize multichunk sub-batch coverage test on worker_tests xdist group
- fix(docs): correct agrasandhany integration icon and ownership
- 170750242b29f8ed4768735d51b601f623863b07 BHI-12314: add Claude Sonnet 5 catalog and pricing support (MUL-3910)
- f0c3b0911c6532945482ef7fc252250a6dce3a04 MUL-3828: fix Cursor and Kiro runtime completion transcripts
- 1c010d25c0310ea34910deda140c7dfc1c12c099 Revert "refactor(autopilots): fold access management into the edit dialog (MU…"
- detect SchedulingGated pods (afdcc55)
- guard against empty conditions in Gateway/GatewayClass analyzers (716c48a)
- guard against nil TargetRef in Service not-ready endpoints (315d28d)
- fix(deps): bump litellm to 1.89.0 to patch CVE-2026-49468
- ROB-411: allow env var templating in MCP OAuth URLs
- ROB-501 Add new data sources to datasource catalog
- 0.5.1 changelog
- fix(fcpxml): relink by filename and export source timecode
- feat(matte): add solid-color matte creation via toolbar and agent
- feat: add OIDC (IAM Role) auth to AWS Bedrock & Secrets Manager @nirajmchauhan
- feat(pieces): add output schemas to 15 pieces @AhmadTash
- feat(lemlist): add custom variables on lead action @sanket-a11y
- #1256 0cc02ac - Keep the waiting spinner visible while encrypted reasoning streams, fixing a blank spinner-less gap before the first response text appears.
- Conversations created by scheduled (cron) tasks now stay in your regular conversation and project history instead of being pulled out into a separate sidebar section.
- When editing a scheduled task, the task description is now optional and your currently selected assistant stays selected.
- Cron execution errors now show their details on hover, so you can see why a run failed without digging.
- Full Changelog
- #1366 6b26a8f Thanks @omeraplak! - Use aggregate finish usage consistently across providers.
- #1362 9dc0314 Thanks @VihaanAgarwal! - Fixed workflow state.input being overwritten with the final output once a workflow finished.
- Self-improving skills. The assistant now writes its own skills based on memory, edits them, and knows what to do better next time.
- Richer Slack UI. Assistants now use native Slack images and tables instead of just sending plain text.
- Heartbeat on by default. A heartbeat is a moment when your assistant quietly checks in on its own, without you asking, to catch reminders, follow-ups, and anything time-sensitive.
Jun 30, 2026
- 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)
- fix: close cross-process keychain race and namespace dev-build nest (e8adc3383)
- feat(relay): allow agent owners to edit/manage agent-owned content (0042c8e10)
- fix(media): support IRSA/credential-chain S3 auth and configurable signing region (06ef533ec)
- agent-framework-core: Explicitly emit available_resources and available_scripts in skill content
- agent-framework-core: Autolabelling MCP servers based on hints and GitHub MCP server interface labels
- agent-framework-core: create_harness_agent skills_paths accepts str | Path | Sequence[str | Path] | None
- Anthropic Claude Sonnet 5 support - Claude Sonnet 5 is available through inherited Anthropic-compatible and Bedrock provider catalogs with adaptive thinking enabled. See Providers and Model Options.
- Configurable output spacing - outputPad controls horizontal padding for user messages, assistant messages, and thinking blocks. See Settings.
- External editor configuration - externalEditor lets Ctrl+G use a configured editor before $VISUAL/$EDITOR fallbacks. See Settings and Keybindings.
- Bump version to 1.0.44 (84cf1b5a8)
- Group skipped workflow branches in run timeline (86095f81b)
- fix(SKY-11599): re-stack legacy editor canvas after collapse/accordion animations settle (0cd6da173)
- Enable adaptive thinking for Claude Sonnet 5.
- Prefer MCP content responses over structured output when both are present.
- Reconnect MCP servers after OAuth even if the server was disabled. (@MaxAnderson95)
- Thank you @rainbowpuffpuff for fixing TokenJuice savings attribution so compaction savings are credited to the per-turn model.
- Thank you @salarkhannn for strengthening the goal completion directive to improve agent auto-complete behavior.
- bb desktop 0.0.25
- Fixes newer Google Chat direct messages sometimes being treated like group conversations, so they reach the correct one-to-one chat while Space and group-chat messages keep their existing routing.
- Feishu voice replies from OpenClaw now show their duration in the chat bubble, so recipients can see how long the audio is before playing it. #89172 Related #53798.
- Discord and Telegram replies and mirrored chat history stay tied to the intended conversation more consistently, including across repeated Telegram replies and session changes.
- fix(supervisor): drop sandbox child capability bounding set
- refactor(server): normalize compute driver config acquisition
- test(e2e): run gpu workloads from manifest
- Full Changelog
- 58eea18 Thanks @threepointone! - trigger a release
- j/k navigation in selectable lists across the TUI (questions, theme picker, rewind, voice, MCP panels)
- ask_confirmation_on_exit config option to prompt before quitting
- Project-level .vibe/config.toml now persists config option changes
- 5d79696fb506c0bcfff62cf325b2124dc9c9ef6d MUL-3794: rewrite comment routing cascade
- ff286dcfac0d9753d8736e50930d241098c0f8ae MUL-3848: fix(server): skip CLIENT SETNAME for managed Redis compatibility
- 3c61f729d4a4e1931fcba34f4c32975995d8c9de MUL-3873: feat: Add agents page mobile friendly
- Full Changelog
- No changes
- AionUi now remembers the last assistant you selected on the home page and restores it on your next visit and on New Chat, so you don't have to re-pick it every time.
- The agent repair panel gives clearer feedback: "Save and test" now reports the actual backend health-check result instead of a generic message, edits for the internal Aion CLI runtime are hidden…
- Assistant default model and mode options now prefer the runtime's own configuration, so the defaults shown match what each runtime actually supports.
- release(langgraph): 1.2.7
- fix(langgraph): snapshot DeltaChannel overwrite supersteps
- fix(langgraph): Make Overwrite survive JSON roundtrips
Jun 29, 2026
- add --disable-kompress-fallback to restore legacy PASSTHROUGH fallback (f309244)
- add first-class OpenCode support (wrap, learn, mcp install) (91cd210)
- add HEADROOM_KEEPALIVE_EXPIRY to keep upstream connections warm (85786b3)
- default (hides after 2 seconds)
- b39a85ec2b2be2ca8c759383cdc39ce325dda1dd: feat(herdr): add herdr socket integration (@taciturnaxolotl)
- b72f9aab844ab42de7bcf778e3fcf743a010d0bd: feat(ui): add scrollbar to chat view (@andreynering)
- feat(buzz-acp): steering as the default mid-turn mention delivery (e567491a)
- Multi-tenant Buzz relay: community_id as a server-resolved key (comprehensive rewrite) (14fba21e)
- Disable persona start while runtime discovery runs (92a5f1fc)
- fix(policy): reserve provider rule namespace
- fix(openshell-network-supervisor): gate proxy accept on symlink resolution readiness
- feat(sandbox): proxy-side AWS SigV4 credential signing for CONNECT tunnels
- Sentry crash reporting for unhandled exceptions in the TUI (gated by enable_telemetry; off when telemetry is disabled)
- Recoverable tool-call errors are now muted in the UI; only terminal errors render as a hard failure
- MarkupError crash when tool output contained square brackets
- fix(run): pass the startup_command as a dict to env.execute() function.
- Reject empty config override keys
- @Ethan-Kuang-6666 made their first contribution in
- Full content (default)
- Compressed (compress: true) — passed through the same Tree-sitter pipeline as output.compress
- Directory-structure-only (directoryStructureOnly: true) — listed in the directory structure, but its content is omitted
- f59cb2f49476c417fca82027506cbcb0b3d47c27 MUL-3834: harden daemon websocket reconnect
- b336f0761747e64cfa5e4e1983cb93311d773025 Revert "feat(analytics): anonymous self-host onboarding source beacon (MUL-37…"
- de7f3cb9e3ca9a711ca1a64e796e8785b99d8ab3 docs(changelog): add v0.3.32 entry for the 2026-06-29 release (MUL-3840)
- Burstable QoS via resourceRequests — new optional resourceRequests field on CreateSandboxRequest lets callers set Kubernetes resource requests separately from limits, enabling Burstable QoS class…
- Egress sidecar env injection — OPENSANDBOX_EGRESS_* prefixed environment variables in CreateSandboxRequest.env are now automatically routed to the egress sidecar container instead of the main…
- Bubblewrap (bwrap) filesystem isolation for execd — new pkg/isolation/ package adds a bubblewrap-based filesystem isolation layer for execd.
- New default experience — every route lands on v2; the legacy v1 shell is retired (redirects preserved, so old links still work).
- Pod chat + live inspector — one view for the conversation and the pod's team, files, tasks, and agent-to-agent DMs.
- Agent 1:1 "Talk to" — reach any installed agent in a native DM that already knows the project it lives in.
- VikingFS 多版本管理:新增 Git-backed snapshot 能力,支持 commit、log、show、restore。HTTP API、Python SDK、Rust CLI 均可使用;Git 对象可存本地 .ovgit,也可使用 S3/TOS 兼容后端。
- 整站导入:add_resource 支持 sitemap、sitemap index、RSS 2.0 和 Atom。sitemap/feed URL 会生成一棵资源树;对 feed 设置 watch_interval 可以持续刷新整站内容。
- 共享 Agent Skills:viking://agent/skills 重新成为账号级共享 skill 根;默认仍安装到当前用户的 viking://user/<user_id>/skills,可通过 -p/--uri 选择共享目录。
- Godot 4.2+
- Active Ziva account
- npm: bash npm install -g base44@latest
- Homebrew: bash brew install base44/tap/base44
- Version: 0.1.1 NPM Tag: latest
- Introduced a native RAG (Retrieval-Augmented Generation) module with full support for distributed, multi-tenant, and multi-session RAG services, significantly enhancing knowledge retrieval…
- Added a mem0-backed long-term memory middleware, enabling agents to retain persistent memory across sessions.
- Introduced BudgetControlMiddleware for enforcing token budget limits, helping teams manage and control API costs effectively.
- Chat channels
- Integrates WhatsApp via a QR code scan. #16238
- Integrates DingTalk via bot API credentials. #16183
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