- openclaw 2026.5.30-beta.1 — Highlights - Agents and CLI-backed runtimes recover more cleanly from interrupted tool calls, stale session bindings, compaction handoffs, and media delivery retries. (#88129, #88136, #88141, #88162,...
- openclaw 2026.5.28 — Highlights - Agent and Codex runtime recovery is steadier: subagents keep cwd/workspace separation, hook context stays prompt-local, session locks release on timeout abort while live OpenClaw locks...
- openclaw 2026.5.28-beta.3 — Highlights - Agent and Codex runtime recovery is steadier: subagents keep cwd/workspace separation, hook context stays prompt-local, session locks release on timeout abort, stale restart...
- openclaw 2026.5.27 — Highlights - Stronger security and content boundaries: group prompt text is kept out of the system prompt, repeated-dot hostnames are normalized, side-effecting command wrappers and unsafe Node...
- feat: support pinning / starring sessions for quick access — Feature Request: Session Pinning / Starring Problem When using OpenClaw across multiple channels (webchat, Telegram, Discord, etc.), the session list grows quickly. It would be very helpful to be...
- fix(feishu): gate bitable tools by tools config — Summary - add channels.feishu.tools.bitable and channels.feishu.tools.base schema/type/default support - gate feishu_bitable_ registration through Feishu tools config - make top-level...
- Memory/QMD: isolate mcporter sidecars per agent — Summary - Isolate QMD mcporter daemon/config state per agent by writing an agent-scoped mcporter config and passing it to mcporter calls. - Preserve the configured mcporter server definition instead...
- fix(agents): quarantine invalid runtime tool schemas — Summary Pruned the runtime tool-schema quarantine branch down to the assistant-turn failure path. This adds a pre-normalization quarantine for active runtime tools whose parameter schema cannot be...
- docs: clarify Android token comments — Summary - Continue the inline-comment pass after PR #88115, focused on useful Android declaration KDoc and tricky contract notes. - Document Android gateway, auth, handler payload, session, chat, UI,...
- fix(configure): migrate stale Codex defaults after OpenAI auth — Summary What problem does this PR solve? - Fixes a configure/auth mismatch where OpenAI ChatGPT/Codex auth could preserve a stale legacy codex/gpt-5.5 default while adding the canonical...
- feat(subagents): forward toolsAllow from sessions_spawn — Summary - Add optional toolsAllow to sessions_spawn for native embedded subagent runs. - Validate it as a string array, preserve explicit [], and reject unsupported ACP or CLI-backed native paths...
- fix(gateway): emit subagent_ended hook for api.runtime.subagent.run() (#59164) — Summary - Plugin SDK subagent runs dispatched via api.runtime.subagent.run() never reached emitSubagentEndedHookOnce, so plugins listening on api.on("subagent_ended", ...) were silent for every...
- refactor: move runtime state to SQLite — Summary This PR reopens the database-first runtime-state refactor after the accidental direct landing in #78595 was reverted from main. It moves active OpenClaw runtime state away from scattered...
- feat: improve MCP operator controls — Summary - add MCP doctor/status/logout CLI controls with OAuth credential status and safer stdio diagnostics - add a dedicated Control UI MCP settings page with server summaries, enablement edits,...
- fix(agents): strip malformed arg-value suffixes — Summary - Strip the malformed terminal > marker from active read/write/edit path args and exec routing args before validation/guard checks. - Preserve payload text fields (content, edit replacements,...
- fix(scripts): timeout crabbox wrapper sanity checks — Summary - add a bounded timeout (5_000ms) to Crabbox wrapper sanity probes (--version, run --help) - enforce SIGKILL on timeout and return a deterministic timeout status for stale/hung binaries -...
- zh-CN localization coverage is still severely incomplete despite PR #18473 being merged — 问题描述 虽然 #18473 / #18508 已经为 Control UI 添加了简体中文（zh-CN）locale 支持，但从日常实际使用来看，中文本地化的覆盖仍然严重不完整。 具体缺失点 1. CLI 命令和输出全部是英文 - openclaw sessions list、openclaw status、openclaw cron list 等所有 CLI 输出全是英文 - --help...
- Doctor: expose UI freshness health findings — Summary - Exposes Control UI protocol freshness as a structured doctor health check for lint and dry-run repair effects. - Keeps real build/rebuild mutation on the existing...
- refactor: extract media and ACP core packages — Summary - Extract media primitives into private @openclaw/media-core and ACP runtime/shared types into private @openclaw/acp-core. - Rewire core, plugin SDK, package-boundary, native/Jiti SDK...
- feat(workboard): add worker dispatch CLI — Summary - Start ready Workboard cards from dispatch by claiming a bounded batch and launching Gateway subagent worker runs with persisted session/run metadata. - Add openclaw workboard CLI plus...
- fix(ios): subscribe to per-session transcripts so group chats update in real time (#80231) — Summary Fixes #80231 — Aight (iOS) group chats never surface assistant replies in real time. The user must exit the chat and re-enter for the next agent message to appear. 1-on-1 chats stream...
- refactor(telegram): keep topic thread mapping plugin-local — Summary - Keep buildThreadAwareOutboundSessionRoute(...) on its existing single canonical threadId API; no new plugin SDK thread-id shape. - Move Telegram topic mapping into Telegram: OpenClaw...
- fix(xiaomi): support MiMo voicedesign TTS — Summary - add Xiaomi MiMo mimo-v2.5-tts-voicedesign to the speech provider model list - omit preset audio.voice for voicedesign requests while preserving existing preset-voice model behavior - add...
- feat: Telegram Guest Mode (来宾模式) for single-bot personal use (Bot API 10.0) — Feature Request: Telegram Guest Mode Support Telegram Bot API 10.0 (May 8, 2026) introduced Guest Mode — bots can be @-mentioned in any group chat without being a member. The bot receives only the...
- fix(memory): respect qmd status timeout and skip checkpoint exports — Summary This PR fixes the QMD vector status timeout path and keeps the checkpoint-related coverage narrow: 1. probeVectorAvailability() now respects the configured memory.qmd.limits.timeoutMs instead...
- fix: user-pinned model falls back to global chain on quota exhaustion — Problem When a user selects a model via \/model \, \modelOverride\ is stored in \sessions.json\ with \modelOverrideSource: \"user\"\. The \resolveEffectiveModelFallbacks\ function returns \[]\ for...
- fix(agents): resolve exact static-catalog models for plugin-harness cold start (#88510) — Summary Fixes #88510. After a gateway restart, a session using a plugin-harness-owned transport (the Codex harness) for openai/gpt-5.3-codex fails with Unknown model: openai-codex/gpt-5.3-codex and...
- feat(agents): add sessions_send handoff acknowledgements — Summary Implements the Phase 1 A2A handoff acknowledgement contract for sessions_send. This builds on the existing handoff proposal work by adding: - stable handoff ids for resolved sessions_send...
- fix(agents): make media generation reply instruction deterministic based on sourceReplyDeliveryMode — Problem The completion prompt for image/video generation was ambiguous: > "Use the current visible-reply contract: if this session requires message-tool replies, call message(action=\"send\")..."...
- fix(doctor): detect stale gateway service version metadata — Summary - Problem: auditGatewayServiceConfig() does not detect stale OPENCLAW_SERVICE_VERSION metadata in gateway service definitions. The gateway start lifecycle path already checks this...
- feat(deepseek): show provider balance in usage status — What - Registers DeepSeek as a provider usage source and wires the DeepSeek provider plugin to resolve API-key usage auth. - Adds a DeepSeek balance fetcher for GET /user/balance, exposed through the...
- fix(coding-agent): strip Claude Code nesting env vars when spawning claude (#57858) — Summary - Coding-agent skill SKILL.md now spawns Claude Code with env -u CLAUDECODE -u CLAUDE_CODE_SESSION -u CLAUDE_CODE_ENTRYPOINT claude ... so a nested OpenClaw-inside-Claude-Code dispatch does...
- fix #88537: [Bug]: tokenjuice - main Codex bash results are not compacted — Summary - Fixes #88537. - Routes Codex-native commandExecution transcript results through the existing codex agent tool-result middleware before mirrored session persistence. - Normalizes native bash...
- OpenClaw repo stats for 2026-05-31 — 375761 stars, 78439 forks, 2426 contributors
- OpenClaw sponsors (116 total) — 116 sponsors supporting OpenClaw
- fix: clarify paginated plugin count (#2393) — fix: clarify paginated plugin count fix: correct shown-count pluralization Agent-Logs-Url: https://github.com/openclaw/clawhub/sessions/32575bd4-9dc7-4b8f-9536-bdc96f008783 Co-authored-by: BunsDev
- docs: explain skill catalog publishing (#2455) — docs: explain skill publishing workflow docs: polish ClawHub publishing guide
- feat(cli): print stored auth token (#2454)
- feat: add reusable skill publish workflow (#2450)
- feat: export skill bundle content for security datasets (#2448)
- fix(cli): report unknown root commands (#2433)
- fix(auth): let admins publish without github age lookup
- Enhance README with self-hosting tips — Added tips for self-hosting OpenClaw Docker image and skills volume management.
- openclaw v2026.5.28 on npm — Multi-channel AI gateway with extensible messaging integrations
- sulpikar2/Qwen3.5-4B-Claude-4.6-Opus-Reasoning-openclaw
- Iter435/Qwen3.5-4B-Claude-4.6-Opus-Reasoning-openclaw
- OpenClaw on DigitalOcean — Deploy your own personal AI assistant with OpenClaw on a DigitalOcean Droplet® server, a powerful open-source platform that runs entirely on your infrastructure. OpenClaw connects to the messaging...
- OpenClaw privilege escalation vulnerability
- A sane but bull case on Clawdbot / OpenClaw
- OpenClaw Creator Spent $1.3M on OpenAI Tokens in 30 Days
- Show HN: DenchClaw – Local CRM on Top of OpenClaw
- The AI Agent Attack Surface Just Exploded ⚠️ — Four months ago AI agents escaped the chat window. Now they are spreading across the internet as remotely reachable execution systems…Continue reading on Medium »
- The expensive, begrudged slide into the garden. — The only game in town with rounded cornersContinue reading on Medium »
- Manus — AI agent bridges thoughts and actions, excelling in work and life tasks like personalized travel, stock analysis, insurance comparisons, and supplier sourcing, autonomously completing tasks and...
- RT @makofy_official: Your AI worker shouldn't cost more than your coffee. OpenClaw on Makofy → $6.99/month per worker. F
- RT @makofy_official: Your AI worker shouldn't cost more than your coffee. OpenClaw on Makofy → $6.99/month per worker. F
- Good. I now have solid, verified data for the thread. Let me also quickly check the current OpenClaw star count to be ac
- RT @OmarShahine: I'm so excited to share that @github Copilot is supported as an agent runtime for @openclaw along side
- 直近一週間のおさるAIマーケティング構築講座の 7days BootCampが熱かった！ 毎日0時までほぼ缶詰で最終日のオフ会がまた内容が濃かった。 講師の ・石田さんからは、OpenClawのナレッジ統合の考え方 ・レオさんからは、U — 直近一週間のおさるAIマーケティング構築講座の 7days BootCampが熱かった！ 毎日0時までほぼ缶詰で最終日のオフ会がまた内容が濃かった。 講師の ・石田さんからは、OpenClawのナレッジ統合の考え方 ・レオさんからは、Utage構築と運用の流れ を教わり、とても学びとなる一日でした。 https://t.co/zLnPUvE7V3
- 还有人在用 OpenClaw 吗？
- @lennysan @openclaw Running multiple OpenClaw agents across sales, home, podcast, and course workflows means lots of per
- @coreyganim An OpenClaw AI C-suite with COO/CMO/CFO/CEO agents will keep reloading the same business context. Telos can
- @degensing Your OpenClaw orchestrator + Claude Code/Codex workers setup is ideal for shared context reuse. Telos can red
- Binário 52% menor no OpenClaw mostra que latência é o teto da IA. Claude 4.8 exige runtimes enxutos para escala real. O
- @uniswap12 @GOATNetwork @ClawUpAI @openclaw AI落地这波红利看来要被这队组合吃到了
- من و بابای OpenClaw باهم در یک قاب 😂 https://t.co/EBChYPU7Jh
- @gubatron You nailed the core issue: the longer the session, the more history gets resent every turn. Telos addresses th
- I tried a telegram group for my Openclaw and it seemed slower and got overwhelming. I feel more productive in the singl
- AI智能体赛道正从概念走向落地，GOAT Network @GOATNetwork 与ClawUp @ClawUpAI 联手打造的生态体系，凭借清晰的落地路径与强劲的增长数据，成为当前极具价值的布局方向。 近日在多伦多举办的 @OpenC
- @balakhonoff @openclaw Yes, will be fixed in today's release.
- Most companies waste 40% of labor on repetitive tasks. An AI agent eliminates that. Keeps the humans for strategy. #AI
- RT @Abdulaziz_Hmadi: أحد أفضل وكلاء الذكاء الاصطناعي، الشهير OpenClaw تنزله على جهازك وتربطه بأدوات الذكاء الاصطناعي مثل
- @oikon48 Claude Code / MCP / Skills / OpenClaw の文脈設計に関心があるなら、Telos はかなり相性が良いです。portable agent context を cache-aware に再利用
- RT @Ayushjhax: Each model analyzes independently, then they argue. Openclaw orchestrates the debate and only fires when