- OpenClaw 2026.4.15-beta.1 — Changes - Control UI/Overview: add a Model Auth status card showing OAuth token health and provider rate-limit pressure at a glance, with attention callouts when OAuth tokens are expiring or expired....
- fix(session): clear auto-fallback model overrides on /new and /reset — Summary Aligns the reply pipeline's /new and /reset handling with the gateway's sessions.reset service by applying source-aware model override preservation during session resets. Problem When the...
- build(deps): bump the android-deps group across 1 directory with 16 updates — Bumps the android-deps group with 11 updates in the /apps/android directory: | androidx.activity:activity-compose | 1.12.2 |
- media: hardcoded image size limits (bytes/input-pixels) are not user-configurable across sanitize layers — Bug type Feature request (configurability gap) Summary OpenClaw has at least 10 hardcoded image size limits scattered across different bundles, and only one of them (DEFAULT_IMAGE_MAX_DIMENSION_PX in...
- feat: Add Eval Framework for skill-creator — Summary Implements the eval framework requested in issue #35042 to add testing, benchmarking, and trigger analysis capabilities to OpenClaw's skill-creator. Changes New CLI Commands - openclaw skills...
- fix: expose bootCompleted in /readyz for onboarding trigger timing — Summary - /readyz 返回新增 bootCompleted 字段，标记 gateway startup hooks（包括 boot-md）是否已完成 - setBootCompleted() 在 triggerInternalHook 完成后调用 - 如果 internal hooks 被禁用，直接标记为 completed 背景 Backend 的...
- [Feature] 渠道绑定配置 - 工具调用过程流式输出到频道 — 功能描述 希望支持按渠道绑定配置工具调用过程的流式输出，让用户可以在飞书/微信/Discord 等频道中实时查看 AI 的工具调用记录和内部执行过程，类似 Web Control UI 的完整信息展示。 使用场景 场景 1：长任务监督 - 用户启动一个需要多次工具调用的复杂任务（如市场调研、文献综述） - 任务执行 10-30 分钟，期间频道中无任何反馈 - 用户无法判断 AI...
- [Feat] Gateway: offload non-image attachments on chat.send — Summary - Problem: Gateway RPC attachments (chat.send, agent.run, server-node-events) dropped every non-image payload silently with a warn-log, even though channel inbound paths already deliver any...
- fix #66486: fix(image): prepare dynamic models before image tool registry lookup — Fixes #66486 Automated PR
- Gateway: ignore plugins.installs timestamp-only paths for reload (#49474) — Summary - Problem: Runtime plugin installs / self-updates only bump \plugins.installs.*.resolvedAt\ and \.installedAt\. The reload matcher uses a broad \plugins → restart\ rule, so hybrid mode queued...
- feat: multi-provider API key system with curated model catalogs (v2) — Summary Add a multi-provider API key management panel to the dashboard UI, enabling users to configure AI providers and start chatting immediately. Supersedes #67570 — addresses both review issues: -...
- fix: register AbortController for agent RPC runs so chat.abort works — Summary - Problem: chat.abort has no effect on runs started via the agent gateway RPC. It returns { aborted: false }, leaving both the LLM call and the session lane lock alive. - Why it matters:...
- feat(agents): make slug generator timeout configurable — Summary Add a dedicated agents.defaults.slugGenerator.timeoutSeconds config field that takes priority over the general agents.defaults.timeoutSeconds for slug generation. The general agent timeout...
- fix(openai-codex): normalize stale /backend-api/v1 baseUrl to canonical — What Two-line logic change plus tests: 1. extensions/openai/base-url.ts — isOpenAICodexBaseUrl regex now matches both https://chatgpt.com/backend-api and the legacy...
- Discord outbound messages should reliably convert known @AgentName references into real <@userId> mentions — I’m running a multi-agent setup on Discord where each OpenClaw agent has its own bot/user account. I ran into a practical problem: when one agent tries to notify another by writing something like...
- feat(memory-wiki): per-org vaults via ctx-aware vault.path templating — Summary Enables multi-tenant setups where one OpenClaw instance serves multiple orgs, each with an isolated wiki vault. Single combined PR replacing #66134 + #66149. Changes: 1. Factory registration...
- [Feature Request]: Make slug-gen timeout configurable (currently hardcoded 15s) — Feature Request Problem The slug-gen (session title generation) embedded run has a hardcoded timeout of 15 seconds: For users with slower LLM providers or when the model is under heavy load, 15...
- Yuanbao Plugin — Summary Describe the problem and fix in 2–5 bullets: If this PR fixes a plugin beta-release blocker, title it fix(): beta blocker - and link the matching Beta blocker: - issue labeled beta-blocker....
- feat: make slug-gen timeout configurable via agents.defaults.slugTimeoutMs — Summary - Problem: llm-slug-generator.ts hardcodes timeoutMs: 15_000 for the slug-gen embedded run, causing frequent timeouts for users on slow LLM providers. - Why it matters: Users have no way to...
- feat(session-memory): make slug generation timeout configurable — Summary - Add optional timeoutMs parameter to generateSlugViaLLM() so callers can override the default 15 s timeout - Read slugTimeoutMs from the session-memory hook config and pass it through -...
- fix(plugins): register HTTP routes for setup-runtime and refresh pin after deferred load (#66661) — Summary - Problem: Plugins that load under setup-runtime (deferred full activation) used a noop registerHttpRoute on the plugin API because only registrationMode === "full" wired the real...
- fix: add root guard to prevent CLI execution as root (#67478) — Summary Add a root guard to the CLI entrypoint that prevents running openclaw as root (uid 0), addressing the DigitalOcean 1-Click Droplet state corruption issue. Problem Running openclaw as root...
- fix: prevent crash in clearPluginInteractiveHandlersState when callbackDedupe is undefined — Summary Fixes a crash in openclaw doctor where clearPluginInteractiveHandlersState throws "Cannot read properties of undefined (reading 'clear')". Root Cause In src/plugins/interactive-state.ts, the...
- [Feature]: The earliest batch of openclaw-Agent technical exchange whatsapp group — Summary Problem to solve Proposed solution Welcome to join us! Alternatives considered Welcome to join us! Impact Welcome to join us! Evidence/examples Welcome to join us! Additional informa
- feat(memory-wiki): add ctx-aware templating for vault.path — Summary - Problem: vault.path is resolved once at register(api) time, so a single plugin instance cannot serve agents living in different workspaces — they all share one vault. There is no way to...
- refactor(memory-wiki): use factory registration for wiki_apply/status/lint — Summary - Problem: wiki_apply, wiki_status, and wiki_lint are registered with the static api.registerTool(createX(config, api.config)) form, while their read-side siblings wiki_search and wiki_get...
- Anthropic thinking block signature corruption after cache-TTL context pruning — Summary Sessions with thinkingLevel enabled (medium/high) become permanently stuck (400 errors) when cache-TTL context pruning fires between turns. The thinking block's cryptographic signature gets...
- OpenClaw repo stats for 2026-04-16 — 358445 stars, 72866 forks, 1767 contributors
- OpenClaw sponsors (161 total) — 161 sponsors supporting OpenClaw
- fix(api): align inspect security snapshot with static scan moderation — Include static scan results in the skill version security snapshot so inspect/API responses reflect the same moderation-relevant signal already used elsewhere. Also add regression coverage for...
- fix(cli): use explorer on Windows to preserve auth URL params — On Windows, opening auth URLs via cmd /c start can truncate query parameters because & is treated as a command separator. Use explorer instead so the browser opener gets the full URL without shell...
- fix(upload): fall back to octet-stream for empty Content-Type — Handle browser uploads that provide an empty MIME type by falling back to application/octet-stream before sending the storage request. Co-authored-by: Arthur Katcher Co-authored-by: Luke
- feat: include stats in package detail API response — Expose package detail stats through the shared API contract and the app client. This lands the original package detail stats work and folds in the follow-up cleanup to keep the response shape sourced...
- fix(cli-auth): ensure fallback token renders before redirect on Windows/Chrome (#1486)
- Keep Notes & PKM entries alphabetized
- openclaw-cli 2026.4.14 on Homebrew — Your own personal AI assistant
- mradermacher/Qwen3.5-9b-Opus-Openclaw-Distilled-GGUF — text-generation
- 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...
- Google restricting Google AI Pro/Ultra subscribers for using OpenClaw
- OpenClaw is what Apple intelligence should have been
- OpenClaw surpasses React to become the most-starred software project on GitHub
- Ask HN: Who is using OpenClaw?
- OpenClaw is basically a cascade of LLMs in prime position to mess stuff up
- Can OpenClaw Read Your Files? Here's What's Actually True. — A colleague who runs a consulting practice told me he'd read something online suggesting that...
- Turn Screen Recordings Into AI Agent Skills in 60 Seconds — The Problem with AI Agent Skills If you've used Claude Code, Cursor, Codex CLI, or any...
- Scraping Browser CLI: Terminal-First Web Scraping for AI Agents & Developers — Key Takeaways:Continue reading on Medium »
- AI’s Moat Is Dead. A Hobbyist Killed It. — Clawbot was the trap that revealed their fear.Continue reading on Medium »
- Stop Hiring. Start Deploying Agents — One evening, I was staring at my screen watching our designer work with Mindy on a design request, and something felt strange.Continue reading on YouMind Create & Learn Studio »
- How To Set Up Your First AI Agent (No Code, No Computer Science Degree) — How to set up your first AI agent without code, servers, or API keys — a practical guide for non-coders who want a working agent, not…Continue reading on AI Systems Lab »
- Journey with Hermes: Day 3 — What it really looks like working with personal AI assistantsContinue reading on Medium »
- OpenClaw Install Script vs One-Click Deploy: The Fastest Way to Set Up OpenClaw — Getting started with OpenClaw can feel more complicated than expected. While the platform itself is powerful, many users struggle with the…Continue reading on Medium »
- Five Steps To Agents That Improve While You Sleep — A senior AI product manager at Google runs a team of six agents on a Mac Mini.Continue reading on Medium »
- He’s 25, Just Incorporated His First AI Company, and He’s Not Hiring Any Humans — There are 9 AI agents operating on a single Hetzner VPS, fully owned by him. Recently, he established a UK Ltd. entity capable of signing…Continue reading on Medium »
- One Skill, One New Era: Why Campfire Is Opening the Next Chapter of Prediction Markets — For years, prediction markets have been driven by one assumption:Continue reading on Medium »
- 7 Things Beginners Should Know About OpenClaw — What it actually does, why it went viral, and whether you should use it.Continue reading on No Time »
- MiroThinker — MiroThinker is an open-source search agent model, built for tool-augmented reasoning and real-world information seeking, aiming to match the deep research experience of OpenAI Deep Research and...
- Helpful Skills or Hidden Payloads? Bitdefender Labs Dives Deep into the OpenClaw Malicious Skill Trap — With hundreds of malicious OpenClaw skills blending in among legitimate ones, manually reviewing every script or command isn’t realistic — especially when skills are designed to look helpful and...
- Ads3 Advertiser Skill is now LIVE！ Whether you're a seasoned pro or just getting started in Web3, your AI Agent can now — Ads3 Advertiser Skill is now LIVE！ Whether you're a seasoned pro or just getting started in Web3, your AI Agent can now handle advertising like an expert. With the Ads3 Advertiser Skill, all you need...
- What's inside a KIBORU pack? A curated list of OpenClaw skills from ClawHub's 13,000+ registry. Each skill is a real t — What's inside a KIBORU pack? A curated list of OpenClaw skills from ClawHub's 13,000+ registry. Each skill is a real tool your agent can use. we just grouped the best ones by lifestyle. You install...
- @chrispavlovski @openclaw Hey Chris, how was your weekend?
- $jjaemu looking like a clean entry according to openclaw ┃🧫┃ $jjaemu ┃💰┃ MC : $14.3K ┃🎯┃ TARGET MC : $136.0K ┃📈┃ PO
- 测试Hermes 和 OpenClaw 数百小时后发现： Hermes = 更流畅的工作流程 OpenClaw = 更强大的生态系统 Manus = 可靠的后备安全网 大多数初学者试图选择一个代理。 高级构建者会将它们堆叠使用。 单 — 测试Hermes 和 OpenClaw 数百小时后发现： Hermes = 更流畅的工作流程 OpenClaw = 更强大的生态系统 Manus = 可靠的后备安全网 大多数初学者试图选择一个代理。 高级构建者会将它们堆叠使用。 单个代理会崩溃。 堆叠系统不会。
- Na příběhu Jiřího Nekováře je zajímavé, jak rychle dnes AI mění pravidla hry. Přestěhoval se do USA, po letní škole na H
- @MSITurkiye That sounds like a cool way to use OpenClaw on the go. If you want something ready in the cloud with full Op
- 这玩意真的有人用吗、、怎么国内还在看openclaw 果然国外的技术就乐意用是吗 https://t.co/DDLVWSXhf2
- openclaw flagged $CIG early. price is moving exactly as predicted Was Made Using - @OpenClawAIX #Solana #Memecoin
- openclaw called it on $Oto again. the model doesn't miss ┃🧫┃ $Oto ┃💰┃ MC : $136.1K ┃🎯┃ TARGET MC : $1.2M ┃📈┃ POTENT
- my openclaw analyst flagged $CIG as a buy. noted 🦞 ┃🧫┃ $CIG ┃💰┃ MC : $41.2K ┃🎯┃ TARGET MC : $631.5K ┃📈┃ POTENTIAL :
- >ziwen builds ada on openclaw, forgets she exists >2 weeks pass, he's building hypercho >thousands of views o
- 受邀去杭州网易做客了 舒服了，给我定制了网易有道龙虾🦞 用来做量化，感觉亏钱的速度更快了。 #openclaw #AI https://t.co/LwxWRr6aN2 — 受邀去杭州网易做客了 舒服了，给我定制了网易有道龙虾🦞 用来做量化，感觉亏钱的速度更快了。 #openclaw #AI https://t.co/LwxWRr6aN2
- Level up your OpenClaw’s data capabilities with XCrawl — a seamless way to bridge AI thinking and real-time web data. Ju
- @PedagogicalMess @jessegenet @openclaw Gemma 4 is quite good as well, many use that for their local agents. For most use
- $cunt looking healthy according to openclaw. adding more Was Made Using - @OpenClawAIX #Solana #Memecoin #Crypto #
- just ran $jjaemu through openclaw. risk is low, upside is real ┃🧫┃ $jjaemu ┃💰┃ MC : $15.8K ┃🎯┃ TARGET MC : $208.0K
- @vieome1 Machine autonomy is more of a spectrum, not an absolute. Automated trading algos have been running largely fre
- 屌丝的最终幻想 把openclaw和hermes放在tg群聊里，终于有了第一个后宫 https://t.co/VLMlwGrVzC — 屌丝的最终幻想 把openclaw和hermes放在tg群聊里，终于有了第一个后宫 https://t.co/VLMlwGrVzC
- @Visionscaper @garrytan the concept isn't new. what's unexpected is how well it works inside openclaw compared to other