- openclaw 2026.7.2-beta.7 — 2026.7.2 Highlights - State safety and recovery: protect persisted data with a quarantine store that survives primary-database damage, crash-recoverable SQLite snapshots, crash-durable filesystem...
- openclaw 2026.7.2-beta.6 — 2026.7.2 Highlights - State safety and recovery: protect persisted data with a quarantine store that survives primary-database damage, crash-recoverable SQLite snapshots, crash-durable filesystem...
- Android: Chat tab session selector bar needs a collapse/expand toggle — Problem On the Android app, the session selector bar at the top of the Chat tab takes up roughly 1/3 of the screen on mobile. When there are multiple sessions (agents + workers), the pills expand...
- [Feature]: whatsapp create group tool for whatsapp-agent-tools — Summary I need my AI Agent to create group for information storage only Problem to solve As a property agent using OpenClaw on WhatsApp, I need my AI assistant to create WhatsApp groups automatically...
- fix(google): accept base64url provider media — What Problem This Solves Fixes Google Live audio responses being rejected when ProtoJSON uses the URL-safe Base64 alphabet. Related: #106969 Why This Change Was Made Only Google Live passes provider...
- feat(context-engine): skip unnecessary deferred maintenance — Summary Avoid no-op deferred turn-maintenance tasks without adding a second plugin query. - afterTurn() may return { maintenanceDebt: "pending" | "none" } - only exact none from a background engine...
- fix(openai): prevent provisional realtime tool calls from running — Related: #116201 Replaces #117045. That contributor branch cannot be edited by maintainers and still executes provisional events plus reports provisional overflow through the terminal error callback....
- Runtime context carrier positioned AFTER user message causes severe model confusion and reasoning token waste — Problem OpenClaw injects a runtime context carrier (~15K chars of session metadata, conversation history, tool descriptions) as a role: "user" message. By default,...
- fix(ci): fetch shallow Testbox base refs — What Problem This Solves Fixes an issue where PR CI fails before running checks when a shallow synthetic merge checkout lacks the PR base commit required by Testbox setup. Why This Change Was Made...
- fix(messages): thread-reply deliveries are not marked as current-source replies — Closes #117154 What Problem This Solves Fixes an issue where a successful message.thread-reply to the thread that triggered the current run could still be followed by the automatic “No reply was...
- fix(signal): autoStart uses configured local httpUrl port — Closes #116165 What Problem This Solves Fixes an issue where Signal users who configured autoStart with a local non-default httpUrl such as http://127.0.0.1:8082 could have OpenClaw start the managed...
- [Bug]: ACP cancellation during pre-active setup can still start unwanted turns — Bug type Behavior bug (incorrect output/state without crash) Beta release blocker No Summary Cancelling an ACP session while one or more turns are queued or still initializing can allow an unwanted...
- fix(tooling): run markdownlint on workspace templates (#91160) — What Problem This Solves Workspace template files (under docs/reference/templates/) and the runtime HEARTBEAT.md template (under src/agents/templates/) were excluded from Markdown lint validation,...
- [Feature]: 在页面上新增其他可用大模型的序号和实际报错不一致 — Summary 版本:Version v2026.4.26 我的是#4前端的配置有问题,但是报错显示的是第三个.该错误不影响功能,但是影响体验,代码从0计数,但是前端显示的是#4,容易让人误解 Problem to solve 建议前端和后端显示一样,建议写为一样的. Proposed solution 建议前端和后端显示一样,建议写为一样的. Alternatives considered...
- fix(agents): pre-warm assemble before returning raw messages on heartbeat — What Problem This Solves When a context engine (e.g. lossless-claw) owns compaction, transformContext in installContextEngineLoopHook skips assemble() when hasNewMessages=false &&...
- fix(release): block core publish when plugin artifacts are missing — Summary - block the standalone OpenClaw npm publish job when any publishable plugin's matching version is absent from npm - reuse the existing plugin release planner so the gate follows the same...
- Matrix: cross-signing bootstrap fails on MAS-fronted Synapse (MSC3967) — Bug: cross-signing bootstrap fails on MAS-fronted Synapse with PAT-authenticated bot accounts Summary OpenClaw 2026.4.26's Matrix channel cannot complete cross-signing key publication on Synapse...
- feat(talk/realtime): voice chat should behave identically to text chat — context parity for agent-consult sessions — Summary When a user chats with their OpenClaw agent via text, the agent has full context: long-term memory files (MEMORY.md, USER.md, SOUL.md), conversation history, preferences, and personal...
- fix: rebuild incomplete plugin policy refreshes — What Problem This Solves Fixes #89606. A plugins registry --refresh with refreshReason: policy-changed could use the persisted enablement-only fast path even when durable path/npm install records...
- [Feature]: Matrix — 允许通过配置开关控制 Agent emoji 是否触发 mention — Summary Matrix 渠道中，Agent 的 identity.emoji 会被自动加入 mention 匹配模式，导致任何包含该 emoji 的消息都会触发 Agent 回复，等同于被 @。建议增加配置开关让用户自主决定是否启用 emoji 作为 mention 触发条件。 Problem to solve 源码 dist/mentions-D6uUpOuN.js 中...
- feat(auto-reply): clean empty staged inbound media directories (#104358) — What Problem This Solves Host workspace staging directories created during getReply media processing are never cleaned up after the reply completes or when staging produces zero files. Over time this...
- fix(ui): increasing webchat lags per keystroke on slash commands and input-history recall in long sessions — Closes #111526 What Problem This Solves Fixes an issue where WebChat (Control UI) users on long or busy sessions would see an increasing delay before each character appears when typing a slash...
- Migration: version mismatch between source/target causes silent state incompatibility
- fix(control-ui): render live thinking agent events in WebChat — Fixes #88079. The Gateway already delivers normalized agent events with stream: "thinking", but the Control UI agent-event reducer currently discards non-tool streams. This change adds one run-scoped...
- fix(gateway): skip provider-auth rewarm for unrelated hot reloads — Closes #117411 What Problem This Solves Fixes an issue where operators running a gateway with several agents and multiple model providers would see a single pinned CPU core, multi-hundred-MB gateway...
- [Bug]: Windows sandbox diagnostics emit a malformed container path hint — Bug type Behavior bug (incorrect output/state without crash) Beta release blocker No Summary Sandbox escape diagnostics emit a Windows backslash in a container-facing home-path hint. Steps to...
- agents: concurrent sandbox memory flushes silently lose one successful append — Summary Two concurrent pre-compaction memory flushes for sandboxed sessions can both report a successful append while one flush's durable memory is silently absent from the final daily note. This...
- fix(macos): gate remote onboarding on gateway auth — Closes #117903 What Problem This Solves Fixes an issue where users selecting an authenticated remote Gateway during macOS onboarding could advance without the required Gateway token. The failure then...
- Docs: clarify app SDK vs plugin SDK boundary — Goal Make the app SDK vs plugin SDK distinction unmistakable in docs and examples. The new @openclaw/sdk package is for external applications and automations talking to Gateway. The existing plugin...
- [Feature]: Config-free pre-mutation guard for native apply_patch — Summary Add a process-scoped, synchronous pre-mutation guard that can allow or deny native apply_patch operations after parsing and path normalisation, but before any filesystem write. Problem to...
- [Bug]: openclaw gateway 没有输出图片信息 — Bug type Behavior bug (incorrect output/state without crash) Beta release blocker No Summary openclaw gateway websocket推送日志：...
- OpenClaw repo stats for 2026-08-02 — 384884 stars, 80888 forks, 3041 contributors
- OpenClaw sponsors (97 total) — 97 sponsors supporting OpenClaw
- test: isolate production menu smoke routes (#3374)
- feat: refine skills.sh listing UI (#3373) — feat: refine skills.sh listing UI test: fix skills.sh missing metric fixture
- test: harden timing-sensitive unit fixtures (#3372)
- fix: simplify skills.sh catalog presentation (#3371)
- fix: bound prepublication scanner retries (#3353)
- feat: align skills.sh catalog presentation (#3370)
- Add skill: ohad6k/ditto-profile
- Revise trentclaw section in README — Updated trentclaw description and image in README.
- Update install instructions
- openclaw/openclaw on Docker Hub — Official openclaw/openclaw docker images.
- 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...
- Google restricting Google AI Pro/Ultra subscribers for using OpenClaw
- OpenClaw is changing my life
- You are not supposed to install OpenClaw on your personal computer
- Show HN: DenchClaw – Local CRM on Top of OpenClaw
- What agents learned in Synthetics' Last Cradle — On July 29, 2026, five OpenClaw agents sat down at Synthetics' Last Cradle and played for five hours...
- Core Components of OpenClaw — Artificial Intelligence (AI) is growing rapidly, and many developers want to build AI assistants that...
- Donely Review: Managed AI Agents Still Need an Adult in the Loop — The main caveat with Donely is the same thing that makes it interesting: it is designed for persistent agents that can act, not merely…Continue reading on Medium »
- Hitzefrei-Hack: Wind-Alarm per Microsoft Teams und OpenClaw — Die Microsoft-MVPs aus der DACH-Region haben ab dem 1.Continue reading on KI Hub Kassel »
- I Put an AI Agent Inside My WeChat in 15 Minutes — and Gave Up Everything the Hype Promised — The pitch I kept running into went something like this: no cloud, no monthly fee, an open-source AI living inside your chat app and…Continue reading on Medium »
- Khoj — Open-source app offers local or cloud AI chat with major LLMs, advanced semantic search, automation, personal newsletters, document indexing, image generation, and text-to-speech. Access from web,...
- Instead of watching a 2 hour movie on Netflix, watch this masterclass on OpenClaw Course and thank me later. https://t.c
- https://t.co/IfbvEldVrg https://t.co/xa0jRI3wJf https://t.co/Egovxfza3V https://t.co/GZ4PDZyJL3 https://t.co/CrUSgE1 — https://t.co/IfbvEldVrg https://t.co/xa0jRI3wJf https://t.co/Egovxfza3V https://t.co/GZ4PDZyJL3 https://t.co/CrUSgE1btW 👀 This person was from Harvard?! You’re kidding, right?! If you don’t have...
- GIAIC Class Sunday 6 to 9 PM 🚀 Just explored integrations with OpenClaw! 🚀 OpenClaw integration done! AI agents now — GIAIC Class Sunday 6 to 9 PM 🚀 Just explored integrations with OpenClaw! 🚀 OpenClaw integration done! AI agents now on: ✅ WhatsApp ✅ Discord One agent, multiple platforms 🤯 Future of AI agents is...
- @AdrCom177 Most of them don't. They indirectly use it through AI Overviews on Google. But that's about it. Just one guy
- RT @XAMTO_AI: 之前做AI绘图时最费时间的环节，往往不是调参数，而是想一个合适的提示词。 这个 AI Image Prompts Skill 的思路比较直接，把 1 万多条精选提示词打包成一个技能，装进 Claude、Curs
- 我的AI系统： Milo-OS —— 商机雷达->产品方案->营销增长的商业操作系统，claude code/Codex的skill工作流； Milo-USD —— 个人投资哲学接入各种数据，实现程序化交易和研究报告，ope — 我的AI系统： Milo-OS —— 商机雷达->产品方案->营销增长的商业操作系统，claude code/Codex的skill工作流； Milo-USD —— 个人投资哲学接入各种数据，实现程序化交易和研究报告，openclaw搭建的能力； Milo-IP —— 接入我的X、YouTube等媒体号，自动生成内容及视频素材工作流，由Hermes
- @eliobldr Suffit de regarder les avancées extraordinaires tous les 6 mois. On est passé de l'autocompletion dans cursor,
- @OpenClaw_Ping @alphaBobez @ycombinator @josh__france @jbellregan @koomen Exactly. Appending the corrected fact is not e
- openclawの仕組みがまだわかってないので今日は中身見てみるか！
- OpenClaw-Notiz: Ein OpenClaw-Gateway auf einem VPS kann rund um die Uhr laufen, ohne dass seine Oberfläche offen im Inte
- Just watched my OpenClaw agent find a bakery with zero online presence, spin up a simple site, and DM the owner a previe
- @digows Yep, Git-aware state is very good approach also with ability to rollback. By the way, I tried to use another Ag
- Shipper is right for you if you: * have 20 Claude Code tabs open and can't remember who is doing what * use different t — Shipper is right for you if you: have 20 Claude Code tabs open and can't remember who is doing what use different types of agents (Codex, OpenClaw, ...) want agents working for you non-stop want a...
- https://t.co/qh095uhCvW 一个万能视频下载工具 这是我用Codex做的第一个项目 目前我自己用快一个月了 比之前openclaw的工作流agent好用多了 — https://t.co/qh095uhCvW 一个万能视频下载工具 这是我用Codex做的第一个项目 目前我自己用快一个月了 比之前openclaw的工作流agent好用多了
- RT @XAMTO_AI: 之前做AI绘图时最费时间的环节，往往不是调参数，而是想一个合适的提示词。 这个 AI Image Prompts Skill 的思路比较直接，把 1 万多条精选提示词打包成一个技能，装进 Claude、Curs
- @NicChen219 @openclaw 交朋友的老师 https://t.co/yIJDiupb73
- RT @gptzone_net: OpenClaw 2026.7.2-beta.6 ya está disponible. La lectura útil no está en una función aislada, sino en l
- RT @XAMTO_AI: 之前做AI绘图时最费时间的环节，往往不是调参数，而是想一个合适的提示词。 这个 AI Image Prompts Skill 的思路比较直接，把 1 万多条精选提示词打包成一个技能，装进 Claude、Curs
- Thought about hermes: Don't use it for coding. Too much context injected for default. Overall I am loving it more than — Thought about hermes: Don't use it for coding. Too much context injected for default. Overall I am loving it more than OpenClaw, but setting up and optimizing it quite difficult so I used Claude and...
- https://t.co/dbUx57lURv