OpenClaw Newsletter - 2026-07-29
OpenClaw Newsletter
Wednesday, July 29, 2026Top Stories
- fix(cron): default isolated agentTurn delivery to none, don't fail run on delivery error — Summary Fixes #90378 Changes the default delivery mode for isolated agentTurn cron jobs from "announce" to "none", and prevents delivery errors from failing the entire cron run. Problem When...
- Add reset family carryover summaries — Behavior or issue addressed Depends on #90239. When a session has been reset/rotated, the active transcript can be clean while the previous reset archive still contains the latest compaction summary....
- fix(workboard): preserve proof history across later card writes — Closes #113309 Related: #110201, #111324 What Problem This Solves Fixes a Workboard data-loss bug where a later comment, claim, status change, or other unrelated mutation could permanently remove...
- fix(gateway): skip pending reply drain on skip-deferral restart — Fixes #89528 Summary openclaw gateway restart --safe --skip-deferral skips the pre-restart active-work deferral check but still waits on the close-stage pending reply drain, because the skipDeferral...
- fix(macos): eliminate subprocess polling for SSH tunnel readiness — Related: #90548 What Problem This Solves Fixes an issue where macOS users connecting to a remote Gateway could incur repeated lsof and ps process launches every 100 ms while the managed SSH tunnel...
- ACP session half-initialization causes permanent ready-check timeout loop — Bug Description When initializeSession() times out during ACP session creation, it leaves a half-initialized session state — the .jsonl session file is created (with header only, ~142 bytes) but the...
- Fix: gateway restart detects npm package version updates — Summary\n\nFix the /restart command so that when the globally-installed npm package has been updated (e.g. npm update -g openclaw), the gateway performs a full process restart instead of a SIGUSR1...
- fix(signal): group allowlist entries silently ignored when group id uses URL-safe base64 — What Problem This Solves Resolves a problem where users configuring channels.signal.groupAllowFrom with a Signal group id copied from a signal.group invite link (or pasted from another tool) would...
- fix(feishu): DM replies fail with error 230101 — Closes #109409 What Problem This Solves Feishu can reject a bot reply sent to a direct-message sender's open_id with HTTP 400 and provider error 230101, even though delivering the same reply to the...
- fix(feishu): re-throw unresolved SecretRef errors from media fallbacks — Summary Re-throw FeishuSecretRefUnavailableError in two catch blocks inside feishuOutbound.sendText (local-image auto-convert path) and feishuOutbound.sendMedia that previously swallowed it silently....
- fix: prevent silent message loss from EmbeddedAttemptSessionTakeoverError — Summary When the OpenAI SDK retried an ECONNRESET internally (default maxRetries=2), the session write lock was released during the retry window. Incoming steering messages could then modify the...
- fix: restore trustworthy end-to-end QA and cross-channel delivery — What Problem This Solves Fixes operator-visible false positives in native QA execution, dropped or unsafe gateway restart recovery, account-crossing or transport-specific channel delivery failures,...
- fix(sessions): deliver agent announcements to direct message sessions — Closes #83046 Related: #83048 What Problem This Solves Fixes an issue where a completed agent-to-agent announcement silently disappears instead of reaching a direct-message conversation when the...
- fix(a2a): resolve direct/dm announce target from session key fallback — Summary A2A announce target resolution may fail for direct-session keys shaped like agent:main::direct: (including dm alias) when the primary session parser path returns null. Impact Direct announce...
- fix(telegram): route poll answers into sessions — Fixes #89573. What Problem This Solves OpenClaw can send Telegram polls, but current main does not route a user's answer from a public/non-anonymous poll back into the originating agent session....
- fix(nextcloud-talk): reject reactions for disabled or unconfigured accounts — What Problem This Solves The Nextcloud Talk react message action dispatched straight to the sender without re-checking the resolved account. A user who disabled their Nextcloud Talk account...
- feat(gateway): manage audio and video attachments end to end — What Problem This Solves Audio and video attachments could lose their managed-media identity and metadata as they moved through Gateway chat, agent replies, transcript persistence, artifacts, and...
- feat(talk): support GPT-Live over gateway relay — Related: #115622 What Problem This Solves GPT-Live worked in browser Talk, but clients using the Gateway realtime relay could not establish a GPT-Live media session. The relay also lacked the...
- [AI-assisted] docs(onboard): document gateway password auth mode
- fix: antml:/mm: namespaced tool-call XML leaks into chat replies (#97750) — Closes #97750 What Problem This Solves Some deployments serve a Claude/MiniMax model through a proxy that intermittently degrades a native tool_use block into plain text. In that degraded case the...
- refactor(qa-channel): reuse shared toErrorObject coercion — What Problem This Solves extensions/qa-channel/src/bus-client.ts carried a private toLintErrorObject(value, fallbackMessage) helper that coerces a non-Error rejection into an Error at the postJson...
- refactor(ui): move Agent Defaults entry from the sidebar into the Agents page — What Problem This Solves Agent Defaults is template configuration touched only a few times per install, yet it held a top-level sidebar slot. With sticky model selection (#115717), the model defaults...
- fix(synology-chat): stop fabricating send message ids — What Problem This Solves Synology Chat outbound sends fabricated a platform message id. After the incoming-webhook sendMessage/sendFileUrl calls resolve with a bare true...
- [Feature]: Decide persisted memory Gateway read contract — Summary Decide the public Gateway RPC contract for read-only persisted memory enumeration before implementing it. This is factored out of #115303 so maintainers can review the sticky API surface...
- docs(agents): restructure root AGENTS.md around product doctrine and deduped policy — What Problem This Solves The root AGENTS.md had grown into ~90% process mechanics with no product-judgment layer, and its highest-cost policies were duplicated: the source-trust/proof-routing...
- fix(setup): show Gemini CLI recovery actions — Closes #115834 What Problem This Solves Fixes an issue where users with Gemini CLI installed would be told they were logged out and shown a dead-end "not auto-tested" row when current Gemini CLI...
- tool-result-truncation can reduce tool outputs to empty, breaking long sessions — Summary During long sessions, tool calls (exec, read, write, session_status, tavily_search, memory_store) begin returning empty results even though the underlying commands succeed. The only recovery...
- feat(zalouser): surface inbound photo attachments via media array — Summary Surfaces inbound photo attachments on the Zalo personal-account channel as native vision content blocks. Previously the zalouser monitor handler dropped Zalo photo metadata, so agents only...
- Tool outputs sometimes render as image attachments and become unreadable to the agent — Summary In long-running / ANSI-heavy tool workflows, tool results can collapse into an image attachment placeholder like (see attached image) from the agent's perspective. Once that happens, the...
- chore(ui): refresh control ui locales — What Problem This Solves Keeps generated Control UI locales synchronized without bypassing protected-branch checks. Why This Change Was Made The Control UI Locale Refresh workflow generated this...
- OpenClaw repo stats for 2026-07-29 — 384465 stars, 80789 forks, 3001 contributors
- OpenClaw sponsors (97 total) — 97 sponsors supporting OpenClaw
- feat(claws): publish CLAW.md prompts (#3262) — feat(claws): publish CLAW.md prompts docs(claws): link prompt bridge PR fix(claws): align prompt package validation test(claws): repin OpenClaw prompt contract test(claws): repin updated OpenClaw...
- fix: add exact and paginated prefix skill discovery (#2969) — fix: add exact and prefix skill search modes docs: document exact and prefix skill search Co-authored-by: Patrick Erichsen
- Add package changelog previews for plugin publishing (#2947) — Add package changelog previews fix: gate package changelog previews * Fix package changelog preview races
- feat: update publish form UI (#3286)
- feat: add audited package hard delete (#3282)
- fix: prevent prepublication retry starvation (#3280)
- fix(ci): accept npm 12 pack --json object shape in release (#3276) — npm 12 returns a package-keyed object from npm pack --json instead of an array. The CLI packages.ts path already dual-parses; the release workflow still assumed an array and would fail packing the...
- feat: add canonical mixed skill search (#3264) — feat: add canonical mixed skill search test: add permanent Test search proof
- Revise trentclaw section in README — Updated trentclaw description and image in README.
- peterjohannmedina/Medina-Qwen3.5-27B-OpenClaw
- mradermacher/Qwen3.5-9b-Opus-Openclaw-Distilled-i1-GGUF — text-generation
- OpenClaw — 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
- Show HN: DenchClaw – Local CRM on Top of OpenClaw
- How to (easily) run a fully local, private AI assistant with OpenClaw and QVAC — What an agent harness is, and what OpenClaw does A language model on its own can only...
- My OpenClaw Agent Keeps Breaking at 3 AM. So I Taught It to Fix Itself. — My OpenClaw agent crashed three Sundays ago. Not a slow degradation — it just stopped. The cron job...
- Core Components of OpenClaw — Artificial Intelligence (AI) is growing rapidly, and many developers want to build AI assistants that...
- I Built a Token-Routing Decision Tree for My OpenClaw Agent. It Cut My Bill and Stopped Three Cron Failures — I burned $87 on a single weekend in March. Not because my OpenClaw agent did anything wrong — because...
- Email Is Not the Universal Agent Protocol: What I Found Testing It — An honest postmortem on my AI email system — the bug is in AgentMail's API, not Gmail forwarding, and the email-as-agent-protocol thesis doesn't hold up.
- I Installed 5 OpenClaw Skills Last Week. Three of Them Replaced My Morning Routine. — There's a moment when you stop thinking of your AI agent as a chatbot and start thinking of it as...
- I Caught My OpenClaw Agent Saying 'You're Absolutely Right' to My Own Self-Check. So I Killed the Self-Check Loop. — A first-person post-mortem on the self-validation loop my OpenClaw agent was hiding from me — and the 40-line fix that broke the cycle without breaking the agent.
- ClawHub Skill Spotlight: Self Improving Agent — AI assistants make the same mistakes over and over. This skill tries to fix that — it captures errors, corrections, and learnings and turns them into persistent
- OpenClaw vs Hermes Agent: Which AI Agent is Better? — A Simple Comparison to Help You Pick the Right Self-Hosted AI Agent for Your Workflow.Continue reading on Medium »
- Maid (AI) — Maid is a cross-platform Flutter app for interfacing with GGUF / llama.cpp models locally, and with Ollama and OpenAI models remotely.
- Apollo AI — Customizable interface for private, on-device AI model chatting, open source and closed source LLM connections, OpenRouter support, custom backend support, and unified access to language models...
- RT @entworacy: 모바일에 직접적으로 접근할 수 있는 Openclaw 느낌의 에이전트인, OpenMinis라는 앱이 있네요. iPhone 기준으로 생각보다 다양한 작업(건강 데이터, 사진, NFC, Visi
- 🤖 𝟭𝟬 𝗔𝗜 𝗦𝗸𝗶𝗹𝗹𝘀 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗧𝗼 𝗞𝗻𝗼𝘄 𝗜𝗻 𝟮𝟬𝟮𝟲 ① 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 ✍️ 🔹 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀: The difference between average output an — 🤖 𝟭𝟬 𝗔𝗜 𝗦𝗸𝗶𝗹𝗹𝘀 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗧𝗼 𝗞𝗻𝗼𝘄 𝗜𝗻 𝟮𝟬𝟮𝟲 ① 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 ✍️ 🔹 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀: The difference between average output and answers you can act on. 💡 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝗶𝘁: Any time you need AI to think like a...
- RT @Chinazhidx: 🚀NEW: #Alibaba open-sourced Qwen-Audio-Agent It doesn’t replace OpenClaw or Hermes. It gives existing A
- 👾 AMH Dashboard v2 is Live 👽 🤘 Today marks a major milestone in the Adaptive Markets Hypothesis (AMH) project. We've m — 👾 AMH Dashboard v2 is Live 👽 🤘 Today marks a major milestone in the Adaptive Markets Hypothesis (AMH) project. We've moved beyond static market analysis to a live quantitative crypto market scanner...
- 【Hermes Agent】いま世界が熱狂しているOpenClaw超え?の自己改善型のAIエージェント。 https://t.co/SjQlnGcBQd 面白そう！
- @gpriorrodrigues Se uma pessoa for ao perfil consegue-se dizer o dia em que o OpenClaw foi configurado para postar por e
- 🤖 𝟭𝟬 𝗔𝗜 𝗦𝗸𝗶𝗹𝗹𝘀 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗧𝗼 𝗞𝗻𝗼𝘄 𝗜𝗻 𝟮𝟬𝟮𝟲 ① 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 ✍️ 🔹 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀: The difference between average output an — 🤖 𝟭𝟬 𝗔𝗜 𝗦𝗸𝗶𝗹𝗹𝘀 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗧𝗼 𝗞𝗻𝗼𝘄 𝗜𝗻 𝟮𝟬𝟮𝟲 ① 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 ✍️ 🔹 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀: The difference between average output and answers you can act on. 💡 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝗶𝘁: Any time you need AI to think like a...
- 60% of the time, the people who thought OpenClaw would be a gamechanger thought Clubhouse would be a gamechanger every s
- @MedusaOnchain I guess it happened because Anthropic did the same and banned Openclaw
- RT @_GateAI: Gate Connect auto-detects Claude Code, Cursor, OpenClaw, Codex, and OpenCode on your machine and routes eac
- The OpenClaw creator asked one question on X and by the next day it had a manifesto, a backlash, and a brand new name: G
- RT @AI_masaou: AIキャラは少し前まで「そろそろ来るトレンド」だと思ってたけど、本格的に来るのはまだ先かもしれないな HermesやOpenClawのような粗い形で一時的にバズることはあると思うけど、本質的にAIキャラが定着
- @tellaho Hey @tellaho , trying to overcome my own reservations and give Buzz a go. I have my openclaw and hermes agent
- RT @JulianGoldieSEO: Instead of watching a 2 hour movie on Netflix, watch this masterclass on OpenClaw Course and thank
- RT @opensourcelab9: 【殿堂】 もう定番なのに、いまだに毎日スターが増え続けているOSSを8選。 数年前の流行りではなく、いま現役で使われている側です。 1. openclaw/openclaw 自分の端末で動く個人用ア — RT @opensourcelab9: 【殿堂】 もう定番なのに、いまだに毎日スターが増え続けているOSSを8選。 数年前の流行りではなく、いま現役で使われている側です。 1. openclaw/openclaw 自分の端末で動く個人用アシスタント。非営利団体が公開で開発して…
- RT @simplifyinAI: Someone just built a self-evolving OpenClaw wrapper and dropped it free on GitHub. It's called MetaCl
- RT @Kyle_Structure: AI builders on X 🤖 If you’re using: • Claude Code • Codex • Cursor • OpenClaw • AI agents • Automa — RT @Kyle_Structure: AI builders on X 🤖 If you’re using: • Claude Code • Codex • Cursor • OpenClaw • AI agents • Automation workflows dro…
- RT @nic_carter: Haven’t heard about openclaw in months
- RT @wesbillman: Buzz Desktop v0.5.0 🐝 Early support for your own harness is here. Try Cursor, OpenCode, Hermes, OpenClaw — RT @wesbillman: Buzz Desktop v0.5.0 🐝 Early support for your own harness is here. Try Cursor, OpenCode, Hermes, OpenClaw, or a custom ACP r…
- AI builders on X 🤖 If you’re using: • Claude Code • Codex • Cursor • OpenClaw • AI agents • Automation workflows drop — AI builders on X 🤖 If you’re using: • Claude Code • Codex • Cursor • OpenClaw • AI agents • Automation workflows drop a 👋 and tell me what you’re building. Always happy to connect with people...
Trending on X
- RT @entworacy: 모바일에 직접적으로 접근할 수 있는 Openclaw 느낌의 에이전트인, OpenMinis라는 앱이 있네요. iPhone 기준으로 생각보다 다양한 작업(건강 데이터, 사진, NFC, Visi
- 🤖 𝟭𝟬 𝗔𝗜 𝗦𝗸𝗶𝗹𝗹𝘀 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗧𝗼 𝗞𝗻𝗼𝘄 𝗜𝗻 𝟮𝟬𝟮𝟲 ① 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 ✍️ 🔹 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀: The difference between average output an — 🤖 𝟭𝟬 𝗔𝗜 𝗦𝗸𝗶𝗹𝗹𝘀 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗧𝗼 𝗞𝗻𝗼𝘄 𝗜𝗻 𝟮𝟬𝟮𝟲 ① 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 ✍️ 🔹 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀: The difference between average output and answers you can act on. 💡 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝗶𝘁: Any time you need AI to think like a...
- RT @Chinazhidx: 🚀NEW: #Alibaba open-sourced Qwen-Audio-Agent It doesn’t replace OpenClaw or Hermes. It gives existing A
- 👾 AMH Dashboard v2 is Live 👽 🤘 Today marks a major milestone in the Adaptive Markets Hypothesis (AMH) project. We've m — 👾 AMH Dashboard v2 is Live 👽 🤘 Today marks a major milestone in the Adaptive Markets Hypothesis (AMH) project. We've moved beyond static market analysis to a live quantitative crypto market scanner...
- 【Hermes Agent】いま世界が熱狂しているOpenClaw超え?の自己改善型のAIエージェント。 https://t.co/SjQlnGcBQd 面白そう！
- @gpriorrodrigues Se uma pessoa for ao perfil consegue-se dizer o dia em que o OpenClaw foi configurado para postar por e
- 🤖 𝟭𝟬 𝗔𝗜 𝗦𝗸𝗶𝗹𝗹𝘀 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗧𝗼 𝗞𝗻𝗼𝘄 𝗜𝗻 𝟮𝟬𝟮𝟲 ① 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 ✍️ 🔹 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀: The difference between average output an — 🤖 𝟭𝟬 𝗔𝗜 𝗦𝗸𝗶𝗹𝗹𝘀 𝗬𝗼𝘂 𝗡𝗲𝗲𝗱 𝗧𝗼 𝗞𝗻𝗼𝘄 𝗜𝗻 𝟮𝟬𝟮𝟲 ① 𝗣𝗿𝗼𝗺𝗽𝘁 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝗶𝗻𝗴 ✍️ 🔹 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗶𝘀: The difference between average output and answers you can act on. 💡 𝗪𝗵𝗲𝗻 𝘁𝗼 𝘂𝘀𝗲 𝗶𝘁: Any time you need AI to think like a...
- 60% of the time, the people who thought OpenClaw would be a gamechanger thought Clubhouse would be a gamechanger every s
- @MedusaOnchain I guess it happened because Anthropic did the same and banned Openclaw
- RT @_GateAI: Gate Connect auto-detects Claude Code, Cursor, OpenClaw, Codex, and OpenCode on your machine and routes eac
- The OpenClaw creator asked one question on X and by the next day it had a manifesto, a backlash, and a brand new name: G
- RT @AI_masaou: AIキャラは少し前まで「そろそろ来るトレンド」だと思ってたけど、本格的に来るのはまだ先かもしれないな HermesやOpenClawのような粗い形で一時的にバズることはあると思うけど、本質的にAIキャラが定着
- @tellaho Hey @tellaho , trying to overcome my own reservations and give Buzz a go. I have my openclaw and hermes agent
- RT @JulianGoldieSEO: Instead of watching a 2 hour movie on Netflix, watch this masterclass on OpenClaw Course and thank
- RT @opensourcelab9: 【殿堂】 もう定番なのに、いまだに毎日スターが増え続けているOSSを8選。 数年前の流行りではなく、いま現役で使われている側です。 1. openclaw/openclaw 自分の端末で動く個人用ア — RT @opensourcelab9: 【殿堂】 もう定番なのに、いまだに毎日スターが増え続けているOSSを8選。 数年前の流行りではなく、いま現役で使われている側です。 1. openclaw/openclaw 自分の端末で動く個人用アシスタント。非営利団体が公開で開発して…
- RT @simplifyinAI: Someone just built a self-evolving OpenClaw wrapper and dropped it free on GitHub. It's called MetaCl
- RT @Kyle_Structure: AI builders on X 🤖 If you’re using: • Claude Code • Codex • Cursor • OpenClaw • AI agents • Automa — RT @Kyle_Structure: AI builders on X 🤖 If you’re using: • Claude Code • Codex • Cursor • OpenClaw • AI agents • Automation workflows dro…
- RT @nic_carter: Haven’t heard about openclaw in months
- RT @wesbillman: Buzz Desktop v0.5.0 🐝 Early support for your own harness is here. Try Cursor, OpenCode, Hermes, OpenClaw — RT @wesbillman: Buzz Desktop v0.5.0 🐝 Early support for your own harness is here. Try Cursor, OpenCode, Hermes, OpenClaw, or a custom ACP r…
- AI builders on X 🤖 If you’re using: • Claude Code • Codex • Cursor • OpenClaw • AI agents • Automation workflows drop — AI builders on X 🤖 If you’re using: • Claude Code • Codex • Cursor • OpenClaw • AI agents • Automation workflows drop a 👋 and tell me what you’re building. Always happy to connect with people...
Releases
- OpenClaw repo stats for 2026-07-29 — 384465 stars, 80789 forks, 3001 contributors
Community
- fix(cron): default isolated agentTurn delivery to none, don't fail run on delivery error — Summary Fixes #90378 Changes the default delivery mode for isolated agentTurn cron jobs from "announce" to "none", and prevents delivery errors from failing the entire cron run. Problem When...
- Add reset family carryover summaries — Behavior or issue addressed Depends on #90239. When a session has been reset/rotated, the active transcript can be clean while the previous reset archive still contains the latest compaction summary....
- fix(workboard): preserve proof history across later card writes — Closes #113309 Related: #110201, #111324 What Problem This Solves Fixes a Workboard data-loss bug where a later comment, claim, status change, or other unrelated mutation could permanently remove...
- fix(gateway): skip pending reply drain on skip-deferral restart — Fixes #89528 Summary openclaw gateway restart --safe --skip-deferral skips the pre-restart active-work deferral check but still waits on the close-stage pending reply drain, because the skipDeferral...
- fix(macos): eliminate subprocess polling for SSH tunnel readiness — Related: #90548 What Problem This Solves Fixes an issue where macOS users connecting to a remote Gateway could incur repeated lsof and ps process launches every 100 ms while the managed SSH tunnel...
- ACP session half-initialization causes permanent ready-check timeout loop — Bug Description When initializeSession() times out during ACP session creation, it leaves a half-initialized session state — the .jsonl session file is created (with header only, ~142 bytes) but the...
- Fix: gateway restart detects npm package version updates — Summary\n\nFix the /restart command so that when the globally-installed npm package has been updated (e.g. npm update -g openclaw), the gateway performs a full process restart instead of a SIGUSR1...
- fix(signal): group allowlist entries silently ignored when group id uses URL-safe base64 — What Problem This Solves Resolves a problem where users configuring channels.signal.groupAllowFrom with a Signal group id copied from a signal.group invite link (or pasted from another tool) would...
- fix(feishu): DM replies fail with error 230101 — Closes #109409 What Problem This Solves Feishu can reject a bot reply sent to a direct-message sender's open_id with HTTP 400 and provider error 230101, even though delivering the same reply to the...
- fix(feishu): re-throw unresolved SecretRef errors from media fallbacks — Summary Re-throw FeishuSecretRefUnavailableError in two catch blocks inside feishuOutbound.sendText (local-image auto-convert path) and feishuOutbound.sendMedia that previously swallowed it silently....
- fix: prevent silent message loss from EmbeddedAttemptSessionTakeoverError — Summary When the OpenAI SDK retried an ECONNRESET internally (default maxRetries=2), the session write lock was released during the retry window. Incoming steering messages could then modify the...
- fix: restore trustworthy end-to-end QA and cross-channel delivery — What Problem This Solves Fixes operator-visible false positives in native QA execution, dropped or unsafe gateway restart recovery, account-crossing or transport-specific channel delivery failures,...
- fix(sessions): deliver agent announcements to direct message sessions — Closes #83046 Related: #83048 What Problem This Solves Fixes an issue where a completed agent-to-agent announcement silently disappears instead of reaching a direct-message conversation when the...
- fix(a2a): resolve direct/dm announce target from session key fallback — Summary A2A announce target resolution may fail for direct-session keys shaped like agent:main::direct: (including dm alias) when the primary session parser path returns null. Impact Direct announce...
- fix(telegram): route poll answers into sessions — Fixes #89573. What Problem This Solves OpenClaw can send Telegram polls, but current main does not route a user's answer from a public/non-anonymous poll back into the originating agent session....
- fix(nextcloud-talk): reject reactions for disabled or unconfigured accounts — What Problem This Solves The Nextcloud Talk react message action dispatched straight to the sender without re-checking the resolved account. A user who disabled their Nextcloud Talk account...
- feat(gateway): manage audio and video attachments end to end — What Problem This Solves Audio and video attachments could lose their managed-media identity and metadata as they moved through Gateway chat, agent replies, transcript persistence, artifacts, and...
- feat(talk): support GPT-Live over gateway relay — Related: #115622 What Problem This Solves GPT-Live worked in browser Talk, but clients using the Gateway realtime relay could not establish a GPT-Live media session. The relay also lacked the...
- [AI-assisted] docs(onboard): document gateway password auth mode
- fix: antml:/mm: namespaced tool-call XML leaks into chat replies (#97750) — Closes #97750 What Problem This Solves Some deployments serve a Claude/MiniMax model through a proxy that intermittently degrades a native tool_use block into plain text. In that degraded case the...
- refactor(qa-channel): reuse shared toErrorObject coercion — What Problem This Solves extensions/qa-channel/src/bus-client.ts carried a private toLintErrorObject(value, fallbackMessage) helper that coerces a non-Error rejection into an Error at the postJson...
- refactor(ui): move Agent Defaults entry from the sidebar into the Agents page — What Problem This Solves Agent Defaults is template configuration touched only a few times per install, yet it held a top-level sidebar slot. With sticky model selection (#115717), the model defaults...
- fix(synology-chat): stop fabricating send message ids — What Problem This Solves Synology Chat outbound sends fabricated a platform message id. After the incoming-webhook sendMessage/sendFileUrl calls resolve with a bare true...
- [Feature]: Decide persisted memory Gateway read contract — Summary Decide the public Gateway RPC contract for read-only persisted memory enumeration before implementing it. This is factored out of #115303 so maintainers can review the sticky API surface...
- docs(agents): restructure root AGENTS.md around product doctrine and deduped policy — What Problem This Solves The root AGENTS.md had grown into ~90% process mechanics with no product-judgment layer, and its highest-cost policies were duplicated: the source-trust/proof-routing...
- fix(setup): show Gemini CLI recovery actions — Closes #115834 What Problem This Solves Fixes an issue where users with Gemini CLI installed would be told they were logged out and shown a dead-end "not auto-tested" row when current Gemini CLI...
- tool-result-truncation can reduce tool outputs to empty, breaking long sessions — Summary During long sessions, tool calls (exec, read, write, session_status, tavily_search, memory_store) begin returning empty results even though the underlying commands succeed. The only recovery...
- feat(zalouser): surface inbound photo attachments via media array — Summary Surfaces inbound photo attachments on the Zalo personal-account channel as native vision content blocks. Previously the zalouser monitor handler dropped Zalo photo metadata, so agents only...
- Tool outputs sometimes render as image attachments and become unreadable to the agent — Summary In long-running / ANSI-heavy tool workflows, tool results can collapse into an image attachment placeholder like (see attached image) from the agent's perspective. Once that happens, the...
- chore(ui): refresh control ui locales — What Problem This Solves Keeps generated Control UI locales synchronized without bypassing protected-branch checks. Why This Change Was Made The Control UI Locale Refresh workflow generated this...
- OpenClaw sponsors (97 total) — 97 sponsors supporting OpenClaw
- feat(claws): publish CLAW.md prompts (#3262) — feat(claws): publish CLAW.md prompts docs(claws): link prompt bridge PR fix(claws): align prompt package validation test(claws): repin OpenClaw prompt contract test(claws): repin updated OpenClaw...
- fix: add exact and paginated prefix skill discovery (#2969) — fix: add exact and prefix skill search modes docs: document exact and prefix skill search Co-authored-by: Patrick Erichsen
- Add package changelog previews for plugin publishing (#2947) — Add package changelog previews fix: gate package changelog previews * Fix package changelog preview races
- feat: update publish form UI (#3286)
- feat: add audited package hard delete (#3282)
- fix: prevent prepublication retry starvation (#3280)
- fix(ci): accept npm 12 pack --json object shape in release (#3276) — npm 12 returns a package-keyed object from npm pack --json instead of an array. The CLI packages.ts path already dual-parses; the release workflow still assumed an array and would fail packing the...
- feat: add canonical mixed skill search (#3264) — feat: add canonical mixed skill search test: add permanent Test search proof
- Revise trentclaw section in README — Updated trentclaw description and image in README.
News
- peterjohannmedina/Medina-Qwen3.5-27B-OpenClaw
- mradermacher/Qwen3.5-9b-Opus-Openclaw-Distilled-i1-GGUF — text-generation
- OpenClaw — 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
- Show HN: DenchClaw – Local CRM on Top of OpenClaw
- How to (easily) run a fully local, private AI assistant with OpenClaw and QVAC — What an agent harness is, and what OpenClaw does A language model on its own can only...
- My OpenClaw Agent Keeps Breaking at 3 AM. So I Taught It to Fix Itself. — My OpenClaw agent crashed three Sundays ago. Not a slow degradation — it just stopped. The cron job...
- Core Components of OpenClaw — Artificial Intelligence (AI) is growing rapidly, and many developers want to build AI assistants that...
- I Built a Token-Routing Decision Tree for My OpenClaw Agent. It Cut My Bill and Stopped Three Cron Failures — I burned $87 on a single weekend in March. Not because my OpenClaw agent did anything wrong — because...
- Email Is Not the Universal Agent Protocol: What I Found Testing It — An honest postmortem on my AI email system — the bug is in AgentMail's API, not Gmail forwarding, and the email-as-agent-protocol thesis doesn't hold up.
- I Installed 5 OpenClaw Skills Last Week. Three of Them Replaced My Morning Routine. — There's a moment when you stop thinking of your AI agent as a chatbot and start thinking of it as...
- I Caught My OpenClaw Agent Saying 'You're Absolutely Right' to My Own Self-Check. So I Killed the Self-Check Loop. — A first-person post-mortem on the self-validation loop my OpenClaw agent was hiding from me — and the 40-line fix that broke the cycle without breaking the agent.
- ClawHub Skill Spotlight: Self Improving Agent — AI assistants make the same mistakes over and over. This skill tries to fix that — it captures errors, corrections, and learnings and turns them into persistent
- OpenClaw vs Hermes Agent: Which AI Agent is Better? — A Simple Comparison to Help You Pick the Right Self-Hosted AI Agent for Your Workflow.Continue reading on Medium »
- Maid (AI) — Maid is a cross-platform Flutter app for interfacing with GGUF / llama.cpp models locally, and with Ollama and OpenAI models remotely.
- Apollo AI — Customizable interface for private, on-device AI model chatting, open source and closed source LLM connections, OpenRouter support, custom backend support, and unified access to language models...
Don't miss what's next. Subscribe to Openclaw Newsletter: