- openclaw 2026.6.2-beta.1 — 2026.6.2 Highlights - Plugin and skill installs now use an operator install policy instead of the old dangerous-code scanner path, with clearer doctor, CLI, ClawHub, and troubleshooting surfaces for...
- openclaw 2026.6.1 — 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 2026.6.1-beta.2 — 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,...
- fix(agents): decode xai and venice tool-call arguments exactly once — Summary On xAI (Grok) and Venice, a tool-call argument string that is meant to contain a literal HTML entity is silently over-decoded before the tool runs and before it is persisted to the...
- fix(feishu): remove overly strict dmPolicy=open wildcard requirement … — PR Description Summary dmPolicy="open" is Feishu's simplest DM access mode — it should allow any sender to message the bot without requiring an explicit allowFrom entry. A Zod schema validation rule...
- fix(infer): load model catalog metadata-only for list/inspect/providers — Summary - Problem: openclaw infer model list, openclaw infer model inspect, and openclaw infer model providers hang indefinitely on 2026.4.27. The grandchild Node process spins at 100% CPU, opens...
- Fix/88033 webchat stop button stuck upstream — PR Description Summary Fix the webchat composer Stop button getting stuck (remaining red/visible) after the assistant finishes replying. The button should revert to Send when the response completes,...
- [AI-assisted] Preserve thread sessions across daily reset by default — Problem Thread-scoped chat sessions currently inherit the global implicit daily reset policy when users do not configure session.resetByType.thread. For long-running Telegram/Slack/Discord topics...
- Improve TUI Gateway queue reliability — Summary - Sends Gateway-backed TUI chat through the Gateway agent RPC so queue mode can handle active runs. - Keeps local TUI submit blocking strict except while finishing context. - Adds optional...
- fix(matrix): restore matrix-js-sdk and crypto-wasm to root dependencies — PR Description Summary Problem: Matrix bridge crash-loops with MODULE_NOT_FOUND on clean npm install -g starting from 2026.5.22. matrix-js-sdk and @matrix-org/matrix-sdk-crypto-wasm are missing from...
- [Core bug] /compact reply silently dropped (replies=0) on all channels in 2026.6.1 — Confirmed: Core dispatch bug — affects ALL channels > Update: Originally reported as a Feishu/Lark plugin issue. After migrating to @openclaw/feishu (built-in), the bug still reproduces with...
- WhatsApp: restart channel on config changes so disabled accounts stop on hot reload (#87951) — Summary - Problem: When channels.whatsapp.accounts..enabled is flipped from true to false, the hot reload detects the config change but does NOT stop the running WhatsApp provider. The Baileys...
- [AI-assisted] Add session history family lookup — Problem When a threaded conversation has already been split across reset transcripts, chat.history and the sessions_history tool only read the currently active transcript. The session store can...
- docs: clarify OpenClaw vs Codex code mode — Summary - add a side-by-side comparison table for OpenClaw code mode vs Codex Code Mode in docs/reference/code-mode.md - add a reverse link from docs/plugins/codex-harness.md so Codex harness readers...
- fix(compaction): log when compaction notice is skipped due to missing block-reply stream — sendCompactionNotice silently returned when onBlockReply was unavailable (e.g. proactive maxHistoryShare-triggered compaction or subagent-completion auto-announce). This made otifyUser: true appear...
- Add channel turn delivery and control-lane diagnostics — Summary This PR adds a small, source-level Agentic Kernel spike for OpenClaw channel turns. The goal is not a rewrite: it makes Telegram/direct-message delivery and control-lane health observable and...
- test: make fs-safe symlink tests compatible with Windows — Replaces the hardcoded Windows skips in the fs-safe symlink-related test cases with dynamic canCreateFileSymlinks and canCreateDirectorySymlinks capability checks. If symlinking is supported by the...
- Unified env var management: .env should be the single source of truth — Problem OpenClaw has multiple layers of environment variable sources with different priorities, which is confusing for users: When a user updates an API key in .env and restarts the gate
- fix(heartbeat): refresh stale Current time line on every helper call (#44993) — Summary appendCronStyleCurrentTimeLine() in src/agents/current-time.ts early-returns when its input already contains Current time:, so heartbeat and cron prompts that flow the previously formatted...
- 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: gate owner-only HTTP tools — Summary What problem does this PR solve? - Restores sender-owner gating for owner-only core tools exposed through Gateway HTTP POST /tools/invoke. - Keeps gateway.tools.allow as an...
- fix(imessage): frame imsg rpc stdout on newline only so U+2028 does not split JSON-RPC responses (#89830) — Summary - iMessage channel head-of-line blocks when a valid imsg rpc --json response carries a raw U+2028 inside a JSON string and the stdout framing splits it before the pending request id resolves....
- Codex/ChatGPT OAuth: 400 (no body) on openai-chatgpt-responses after upgrade to 2026.6.1 (Windows)
- fix #69443: [Bug] Subagent RPC callback to WeChat session key routed to main session instead — Summary - Fixes #69443. - Treat a full agent-scoped to value, such as agent:main:openclaw-weixin:direct:o9cq802hhmfc@im.wechat, as the requested agent session key when no explicit sessionKey or...
- test: detect file symlink support dynamically in json-file tests — Replaces the hardcoded Windows skip in JSON file helper tests with a dynamic file symlink capability check. If file symlinks are supported by the environment, the tests execute. Otherwise, they skip...
- fix(memory): move local embeddings to llama.cpp provider — Ignore the huge diff, this is just a proposal for @vincentkoc Summary Local memory embeddings could break because the core package no longer owns node-llama-cpp. This moves the local llama.cpp...
- fix(agents): guard system prompt tool schema stats — Summary - guard system prompt report generation against throwing tool parameters getters - guard tool schema property counting against throwing properties getters - preserve the existing...
- Idr msteams adaptive card tables — Summary What problem does this PR solve? Microsoft Teams renders markdown tables as plain-text code blocks ( -wrapped ASCII tables), which are hard to read on mobile, lose column alignment in narrow...
- feat(feishu): bot-to-bot conversation support with @mention handling — Summary Lets the Feishu agent take part in bot-to-bot conversations and makes outbound @mentions correct and robust. In a Feishu group a message only notifies the entities it explicitly @mentions, so...
- fix(trajectory): guard tool schema capture — Summary - guard trajectory tool-schema capture against throwing tool parameters getters - guard trajectory payload limiting against throwing nested object field getters - record bounded truncation...
- fix(doctor): archive conflicting plugin install index — Summary - Problem: legacy plugins/installs.json stayed in place when shared SQLite plugin install metadata conflicted with the legacy JSON source, so every later command could re-enter the state...
- OpenClaw repo stats for 2026-06-04 — 376718 stars, 78719 forks, 2478 contributors
- OpenClaw sponsors (112 total) — 112 sponsors supporting OpenClaw
- fix(httpApiV1): surface 400-class delete validation errors instead of opaque 500 (#2496) — Summary: - The branch narrows the soft-delete bad-request substring whitelist from any reserved message to the ClawHu ... rvation phrase and adds regression tests for the intended 400 path and an...
- fix(httpApiV1): surface 400-class delete validation errors instead of opaque 500 (#2488) — Merged via squash. Prepared head SHA: cc6ac9d31a70c7911ab1579fe56bc23b7ee30f4f Co-authored-by: momothemage Co-authored-by: momothemage Reviewed-by: @momothemage
- feat: add clawhub scan command (#2479) — docs: design clawhub scan command docs: plan clawhub scan command feat: add clawhub scan command fix: prune ephemeral scan uploads * fix: avoid scan route slug collision
- feat: show plugin download counts (#2484)
- fix: gate local Codex workers (#2472)
- fix: use GitHub App auth for GitHub account lookups
- fix: harden GitHub account age lookup
- update coderabbit
- update tools section
- Add skill: brewpage-publish
- openclaw v2026.6.1 on npm — Multi-channel AI gateway with extensible messaging integrations
- openclaw-cli 2026.5.28 on Homebrew — Your own personal AI assistant
- mradermacher/Medina-Qwen3.5-27B-OpenClaw-Uncensored-i1-GGUF
- 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...
- Tell HN: Anthropic no longer allowing Claude Code subscriptions to use OpenClaw
- OpenClaw is a security nightmare dressed up as a daydream
- Show HN: DenchClaw – Local CRM on Top of OpenClaw
- Openclaw on the cloud, dos and don'ts — running openclaw or hermes locally sounds nice, but how about running on the cloud?
- I Gave OpenClaw a Voice and It Ordered Me Dinner — How I wired OpenClaw to a microphone, a Murf Falcon voice, and a Swiggy account — and what I learned building a real-time voice agent in 800 lines of TypeScript.
- Cloud-Native Agent Army Command Layer for OpenClaw — Originally published at https://clawbud.ai/blog/cloud-native-agent-army-command-layer ClawBud is...
- NemoClaw for the Enterprise: Policy Engineering (Part 4) — Your agent can talk. Your agent can listen. Now decide exactly what it’s allowed to do with the right...
- From Clawbot To OpenClaw: How 1.5 Million Autonomous AI Agents Took The World By Storm
- J’ai fermé la porte à clé. Sauf que les meubles n’étaient pas dans la bonne pièce. — La suite de mes aventures avec ma grand-mère IA, un agent qui s’appelle Le Goat, et ce que « sécurité » veux dire.Continue reading on Medium »
- I Tested MiniMax M3 as a Coding Agent Through OpenClaw — It did not write perfect code on the first try. But once I added a real build-and-test loop, it became surprisingly useful.Continue reading on Medium »
- The Ultimate Guide to Building Your Own Personal AI Assistant with OpenClaw: From Concept to… — In an era where Artificial Intelligence (AI) has become an inseparable part of our daily digital lives, most users are still limited to…Continue reading on Medium »
- Microsoft Named Its New Assistant Scout. The Part Worth Reading Is the Audit Log — Two days ago NVIDIA sold me an agent that lives on my desk. At Build, Microsoft sold me one that lives in its cloud, and the gap between…Continue reading on Artificial Intelligence in Plain...
- Prompt Injection in Production: The Patterns That Defend Real Systems — The injection that does not look like an injectionContinue reading on Medium »
- OpenClaw vs Hermes: The Agent Comparison That’s About More Than Features — Most comparisons miss the point. Here’s the distinction that actually changes the recommendation.Continue reading on Medium »
- Thunderbolt — Open-source AI platform for deploying language models across desktop and mobile systems, features model selection, on-premises setup, enterprise-level encryption, advanced security auditing, and...
- 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...
- RT @socialwithaayan: 🚨 BREAKING: A university lab just shipped a personal AI agent in 4,000 lines of Python. OpenClaw h
- RT @openclaw: Agents should learn repeated work, but not by silently rewriting future runs. Skill Workshop turns reusab
- OpenClaw 之后最让人惊艳的产品，细节满满 https://t.co/2VeL0yjr1c
- RT @cobie: In January I asked OpenClaw to send 50,000 small invoices to Fortune500 companies every day. Through experi
- Are you still running OpenClaw?
- RT @msdev: Watch @shanselman and Samantha Song demo OpenClaw on Windows at #MicrosoftBuild. In the full walkthrough, y
- RT @ankitkr0: i didn't try openclaw at all but i just tried hermes by @NousResearch and i think this is the best AI expe
- J’ai installé hermes et honnêtement comme openclaw ça vient t’ajouter de la friction à tes agents et surtout ils sont pa
- @JackW1982 @openclaw 我推荐你使用 DeepSeek-v4-flash 或 Pro。
- @Asteri_eth I switched from OpenClaw to Hermes a week ago, and I actually like it more Only thing is token usage feels
- @Teknium @openclaw Great to see even @openclaw acknowledging and appreciating the self learning capability of Hermes Age
- Here's the truth people aren't ready to hear: AI agents don't need to be faster. They need a more performant memory. A
- Hermes AI SEO Workspace: New FREE Mission Control! Hermes Workspace just turned AI SEO into a mission control.⚡ The pr
- 「OpenClaw」がWindowsノードにネイティブ対応、もうペンギンは不要 - 窓の杜 https://t.co/S83nInJayv
- @trikcode I tried to use scout, but it needs GitHub business and Microsoft 365 with copilot accounts. Not very accessibl
- I have never used Claude or OpenClaw, ChatGPT, or any other touted AI to directly write into my codebase. Just don't tru
- Is anyone still using OpenClaw?
- On June 1, GitHub Copilot went usage-based. Coding agents are now a metered cloud cost, not a flat license. One agent s
- @Youssofal_ "Losses", losses to what? You are getting happy over a meaningless benchmark with no real word relevance tak