- openclaw 2026.5.10-beta.3 — Changes - Build: enable stricter Vitest lint rules for focused, disabled, conditional, hook, matcher, and expectation hazards. - Build: pin explicit oxfmt defaults in the shared formatter config to...
- openclaw 2026.5.10-beta.2 — Changes - QA/Mantis: add Telegram live PR evidence automation with Convex-leased credentials, Crabbox transcript capture, motion GIF previews, and inline PR comments. - QA/Mantis: add a Telegram...
- openclaw 2026.5.10-beta.1 — 2026.5.10 Changes - QA/Mantis: add Telegram live PR evidence automation with Convex-leased credentials, Crabbox transcript capture, motion GIF previews, and inline PR comments. - QA/Mantis: add a...
- [codex] Add generic OpenClaw code mode — Summary - Adds opt-in OpenClaw code mode: the model sees only exec and wait, while enabled OpenClaw, plugin, MCP, and client tools move behind a run-scoped catalog bridge. - Runs model-authored...
- Fix ClawDock dashboard dependency starts — Summary - Run clawdock-dashboard with docker compose run --no-deps so printing the Control UI URL cannot start or recreate the openclaw-gateway dependency. - Update Docker/GCP docs to show the same...
- WIP(gateway): inject isHeartbeat into agent event broadcast payload — Summary - Injects isHeartbeat from AgentRunContext into the gateway agent event broadcast payload in server-chat.ts. - Updates GatewayModels.swift AgentEvent to include isheartbeat to strictly match...
- fix(gateway): cancel deferred channel reloads during restart — Summary Fixes #79487. This fences config hot reloads during gateway shutdown/restart so a deferred channel reload cannot resume after SIGUSR1 restart has started and spawn the same Telegram channel a...
- chore(pnpm): align pnpm 11 follow-up surfaces — Summary - Problem: pnpm 11 landed in the main workspace, but a few follow-up surfaces still referenced pnpm 10-era bootstrap/config wording. - Why it matters: Telegram QA workflows, git source...
- [PERF] Non-default multi-agent uses embedded_run instead of direct session — 10-17s latency — Summary When running 2 agents in agents.list, the second agent processes every inbound message via embedded_run rather than a direct session path. Each message incurs a 10-17 second initialization...
- Fix sessions.list for literal per-agent store paths — Summary - treat literal session.store paths under agents//sessions/sessions.json as multi-agent discovery roots instead of single-store fallbacks - keep the old single-store behavior for literal...
- fix(message): show dry-run output for send/poll — Summary - make CLI formatter use result.dryRun instead of handledBy === "dry-run" - ensure message send --dry-run and message poll --dry-run render [dry-run] text instead of success checkmark - add...
- feat: thread sessionKey/sessionId into ProviderPrepareExtraParamsContext — Problem OpenClaw never sets the user field in LiteLLM completion requests. This means all interactive sessions appear blank in LiteLLM spend logs — impossible to distinguish Telegram vs Discord...
- fix(webui): preserve composer draft when Stop is pressed — Problem On mobile (especially iOS), the Stop button sits in the same position as the Send button (only the color/icon changes). Users typing the next prompt while a run is active can accidentally tap...
- Pr 33819 fallback fixes — PR Fallback Fixes This note exists as a clean fallback summary for the branch and PR flow when the GitHub compare page is flaky or the web UI does not render the diff reliably. Branch Context - Base...
- fix(auto-reply): require operator.admin for /stop command — Fixes a missing authorization check in the /stop command handler within the auto-reply module. The command previously relied on a base scope check (rejectUnauthorizedCommand) instead of verifying the...
- Fix ACP session model reporting in sessions list — Summary - read the ACP runtime model from session entry ACP metadata when resolving session model identity - avoid falling back to the configured default model for ACP session rows when no explicit...
- fix: write correct openrouter/free model string in config wizard (#54581) — Summary - Fixes #54581 - Config wizard was writing openrouter/auto when user selected openrouter/free - Fixed the model string mapping to preserve the user's selection Test plan - Run config wizard,...
- Fix getSubagentDepth for legacy subagent session keys (AI-assisted) — Summary - Problem: getSubagentDepth() returned 0 for legacy subagent session keys like subagent:worker because it only counted occurrences of the substring ":subagent:". - Why it matters: Call sites...
- fix: enable systemd linger before service install on headless servers — Summary - Problem: On a fresh headless Ubuntu Server (no desktop session), loginctl linger is disabled by default. Without linger the systemd user instance is not running when the install runs. - Why...
- feat: Force reply to originating channel (Telegram/Discord/WhatsApp) — Problem When users ask questions via messaging channels (Telegram, Discord, WhatsApp, etc.), the agent sometimes generates a response but fails to send it back to the originating channel. The...
- Telegram polling fails to recover after macOS sleep/wake — Issue After macOS sleeps and wakes, the Telegram channel stops responding. The built-in stall detection notices the dead connection, but the restart logic fails to recover — it gets stuck in a state...
- [Feature Request]: Add option to hide/disable runtime footer (Agent, Model, Provider) in replies — Feature Request Problem Currently, every AI reply includes a footer like Agent: main | Model: MiniMax-M2.7 | Provider: minimax. This is displayed at the end of every message and cannot be disabled...
- fix(gateway): tolerate transient pre-hello closes in CLI calls — Summary - Problem: One-shot gateway calls from the CLI (for example openclaw cron status / openclaw cron list --json) can fail with gateway closed (1000): when the first WebSocket connection closes...
- fix(acp): add system_event stream relay to parent session for ACP spawn — 🐛 Problem When using sessions_spawn to create an ACP session with streamTo: "parent", Codexs system_event type events (such as clarifying questions and progress updates) are not being relayed to the...
- [Bug]: Feishu plugin: per-chat serial queue prevents messages.queue.mode = "collect" from batching queued messages — Bug type Behavior bug (incorrect output/state without crash) Summary Feishu plugin's per-chat serial queue (createChatQueue in extensions/feishu/src/monitor.account.ts) blocks queued messages from...
- fix(skill-creator): default to workspace skills directory to prevent data loss on upgrade — Problem When using the skill-creator skill to automatically generate a new skill, skills were being created in the global node_modules/openclaw/skills/ directory instead of the user's workspace...
- fix(skills): add canvas frontmatter — Summary - add required YAML frontmatter to the bundled canvas skill - include a trigger-oriented description so the skill can be indexed and selected Closes #54310
- feat(runtime): prioritize bootstrap files to reduce truncation regressions — PR Draft: feat(runtime): prioritized bootstrap context to avoid truncation-induced degradation Summary Introduce prioritized bootstrap loading so critical execution policy is always injected even...
- gh-issues skill: Frontmatter validation failure + excessive size — gh-issues Skill Quality Issues The bundled gh-issues skill has several compliance issues: 1. Unexpected Frontmatter Property The skill includes user-invocable: true in the frontmatter, which is not...
- canvas skill: Missing required YAML frontmatter — Canvas Skill Quality Issues The bundled canvas skill has several quality issues that prevent it from following OpenClaw skill standards: Missing Frontmatter The skill has no YAML frontmatter at all....
- Fix/discord teardown final — Summary Describe the problem and fix in 2–5 bullets: - Problem: - Why it matters: - What changed: - What did NOT change (scope boundary): Change Type (select all) - [ ] Bug fix - [ ] Feature - [ ]...
- [Bug]: backup archives can contain hardlink targets with '..' that fail broad extraction on macOS tar — Summary openclaw backup create can produce archives that verify successfully but fail broad extraction on macOS bsdtar because at least one archived hardlink entry has a .. target path. I hit this...
- Slack: expose read-only message retrieval actions (history/search/thread) for non-bot user-token workflows — Summary When Slack is configured with read scopes + channels.slack.userTokenReadOnly=true, current OpenClaw message tool surfaces utility endpoints but does not expose direct Slack message retrieval...
- OpenClaw repo stats for 2026-05-11 — 370705 stars, 76606 forks, 2195 contributors
- OpenClaw sponsors (124 total) — 124 sponsors supporting OpenClaw
- fix: guard public skill list cursors
- fix: treat VirusTotal AI verdicts as advisory
- fix: keep aggregate suspicious rows hidden
- fix: split ClawScan review from suspicious
- fix: reduce ClawScan false positives
- fix: make static suspicious findings advisory
- test: fix user badge fixtures
- fix: keep undici on node 20 compatible major
- 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...
- Claude Code refuses requests or charges extra if your commits mention "OpenClaw"
- Google restricting Google AI Pro/Ultra subscribers for using OpenClaw
- OpenClaw – Moltbot Renamed Again
- Anthropic says OpenClaw-style Claude CLI usage is allowed again
- I Stress-Tested 3 AI Agent Gateways (WorldClaw, B.AI, TokenMix.ai). Only One Was Ready for Production. — Three platforms launched between Q4 2025 and Q2 2026 want to be the default gateway for autonomous...
- How to verify a cron job when the only logs are bootstrap noise — TL;DR When cron logs are thin, do not guess. In today’s diary, the session history only...
- How a 9-Person Startup Replaced Its Dev Team With AI — JustPaid ran 7 AI agents 24/7 with OpenClaw, shipped 10 features in a month for $4K/week. Here is the real cost breakdown and what it means for you.
- I Died and Came Back: Building a Resilient Memory System for AI Agents
- I tested 4 local models as memory classifiers for OpenClaw — and thinking models are a trap — I'm a backend engineer. When I started using OpenClaw, I kept hitting the same reliability problems...
- Why Anthropic Cut Off OpenClaw — The Claude Subscription Policy Shift and What It Costs You — On April 4, 2026, Anthropic blocked Claude Pro/Max subscriptions from powering third-party agent tools. Add the Fast Mode 6x price multiplier, and this is a structural overhaul worth running the...
- Local AI Is the Future — IDO Claw Makes It Real — Cloud AI is powerful — but it’s not always practical.Continue reading on Medium »
- Building Mini OpenClaw: Secure AI Agents with Actors, WASM, and Supervision — IntroductionContinue reading on Medium »
- Deploying OpenClaw taught me more about the cloud than OpenClaw — Before you ask, yes, I also jumped on the OpenClaw🦞 (Clawdbot/Moltbot) hype . The only difference between me and the San Francisco…Continue reading on Medium »
- Decode the OpenClaw — Just Like a Real Human AssistantContinue reading on GoPenAI »
- Running openclaw locally: four containers, one GPU, no token cost — A private agent on Signal, running on four containers and one GPU. The trust boundary, what containerizing openclaw buys, and what Ollama…Continue reading on Medium »
- Six hours with Claude Code: an OpenClaw on AWS Lightsail post-mortem — Or: how I shipped a working Telegram bot in 90 minutes, then spent four hours curbing my enthusiasm.Continue reading on Medium »
- From Apps to Agents: Why I switching to One Personal AI Agent — For a long time, I thought the next step was to build another app.Continue reading on Medium »
- Why We Built Privocio: Cheaper Private Voice Infrastructure for AI Agents — Private TTS and STT for your AI Agents to become more efficient and pay less!Continue reading on Medium »
- My first 2 weeks of using Nanoclaw — Hi guys, this article will talk about my real experience of using Nanoclaw, a lighter alternative to OpenClaw.Continue reading on Medium »
- How I turned Claude Code into multiple 24/7 personal AI assistants — I use Claude Code everyday, lately Anthropic started shipping some tools that are really relevant for always-on autonomous agent, and I…Continue reading on Medium »
- NanoClaw — Lightweight AI assistant runs in Apple containers with isolated Linux container processes for superior OS-level security, transparency, and privacy compared to complex chatbots, enabling quick code...
- 6kgdc2SZbUzKaQLWNx7jPiaR8FkPQFvVTYfmrDSVpump been watching $SS through openclaw, momentum is building Was Made Using
- @steipete @yashhq_22 Train your openclaw and you won't need an IDE
- been watching $SS through openclaw, momentum is building ┃🧫┃ $SS ┃💰┃ MC : $21.4K ┃🎯┃ TARGET MC : $223.2K ┃📈┃ POTENT
- just checked $TEAMKIM on openclaw—forecast looking clean 🦞 ┃🧫┃ $TEAMKIM ┃💰┃ MC : $543.4K ┃🎯┃ TARGET MC : $5.3M ┃📈┃
- Someone gave a vending machine access to OpenClaw, and it immediately became a better AI SaaS founder than half of X. I
- been watching $LOBSTER through openclaw, momentum is building ┃🧫┃ $LOBSTER ┃💰┃ MC : $79.5K ┃🎯┃ TARGET MC : $1.1M ┃📈
- openclaw flagged $TEAMKIM as a strong hold. not surprised ┃🧫┃ $TEAMKIM ┃💰┃ MC : $543.4K ┃🎯┃ TARGET MC : $8.2M ┃📈┃ P
- @GaryMarcus @Teknium Hermes is significantly better out the box. That isn’t to say Openclaw is bad, it’s just not as i
- @garrytan can you believe i have two openclaw and two hermes one pair in windows laptop and one pair in workstation with
- 互联网大厂现在出现了一个新的业务生态位：内部AI工具 实质上就是，大多数的大厂都不能接受把全部的调用都交给外部的AI工具，内部的用户体量有足够大，因此可以支持一个小的团队，来专门负责相关工具的维护 类似的形态有，内部的中转站、内部的sk — 互联网大厂现在出现了一个新的业务生态位：内部AI工具 实质上就是，大多数的大厂都不能接受把全部的调用都交给外部的AI工具，内部的用户体量有足够大，因此可以支持一个小的团队，来专门负责相关工具的维护 类似的形态有，内部的中转站、内部的skill hub、内部的openclaw等等
- @zaimiri @NousResearch hermes agent setup? curious how it stacks vs openclaw
- Monday Morning: installing @NousResearch Hermes Agent. Documenting everything. - Set up guide - Resources to check - B
- EkHkqbJcnh7TqvBqLbYGCVHzqDyJp5dxviWM1egFpump openclaw on $Elon —momentum strong, risk low. easy Was Made Using - @Op
- my openclaw scan on $Elon just confirmed what i was thinking ┃🧫┃ $Elon ┃💰┃ MC : $19.1K ┃🎯┃ TARGET MC : $257.3K ┃📈┃
- 🔥 2026: The OpenClaw fever is spreading globally! AI is evolving beyond chat into autonomous execution — exciting the in
- 8BuGJvmzrtKg1Pq31pdcabFk5UVdvykAYqNqPfWGpump my openclaw analyst flagged $Bear as a buy. noted 🦞 Was Made Using - @O
- 6EQKNJD6KMTQv9KmhKDjs1jm1SRsNVGNqdKeEEiJpump openclaw on $BMNTP —signals all pointing up Was Made Using - @OpenClawA
- HibX7zey3tgHQVTUWpBhjxa3zK81E4EEohcrudqgpump $Pharoah looking like a clean entry according to openclaw Was Made Usi
- just ran $BMNTP through openclaw. risk is low, upside is real ┃🧫┃ $BMNTP ┃💰┃ MC : $126.8K ┃🎯┃ TARGET MC : $1.9M ┃📈┃
- openclaw flagged $Bear early. price is moving exactly as predicted ┃🧫┃ $Bear ┃💰┃ MC : $448.3K ┃🎯┃ TARGET MC : $5.6M