OpenClaw Newsletter - 2026-06-11
OpenClaw Newsletter
Thursday, June 11, 2026Top Stories
- OpenClaw 2026.6.6-beta.1 — 2026.6.6 Highlights - Security boundaries are substantially tighter across transcripts, sandbox binds, host environment inheritance, MCP stdio, Codex HTTP access, native search policy, elevated...
- #92109: [Bug]: EmbeddedAttemptSessionTakeoverError caused by Btrfs ctimeNs instability — Summary Fix the session-file fence so Btrfs ctimeNs-only metadata drift does not falsely trip EmbeddedAttemptSessionTakeoverError during cron/subagent prompt-lock release. The maintained branch keeps...
- [Bug]: Telegram isolated ingress deletes spooled updates when the turn fails — errors are swallowed so handleUpdate never rejects (silent message loss) — Summary In Telegram isolated polling ingress mode (default-on), a spooled inbound update is permanently deleted even when processing it failed, because the spool consumer keys deletion on...
- fix(agents): handle "Not supported model" error format in isModelNotFoundErrorMessage — Summary Fixes #92118 isModelNotFoundErrorMessage() in src/agents/live-model-errors.ts did not recognize the "Not supported model " error format returned by some OpenAI-compatible provider gateways...
- Bug: WebChat streaming display shows tool calls/results out of order — requires F5 refresh to correct — Environment - OpenClaw Gateway: 2026.6.5 - Browser: Chrome Problem In the OpenClaw WebChat UI, the real-time streaming display of a multi-step agent turn does not maintain correct ordering. A typical...
- isModelNotFoundErrorMessage does not match "Not supported model" error format — Description isModelNotFoundErrorMessage() in src/agents/live-model-errors.ts does not recognize the error format "Not supported model " returned by some OpenAI-compatible providers (e.g. internal...
- Flaky / failing: simple-completion-runtime > can preserve asynchronous provider model discovery (introduced 25ca39e876) — Flaky / failing test on main: simple-completion-runtime.test.ts > prepareSimpleCompletionModel > can preserve asynchronous provider model discovery Summary The test added in commit 25ca39e876...
- [Bug]: Collapsed sidebar + window resize breaks chat composer input area layout — Summary 折叠导航侧边栏（.shell--nav-collapsed）后左右拉伸/缩放浏览器窗口，WebChat 界面底部输入框（compose 区域）左侧的控制元素无法正确收缩，导致输入框左侧布局错位/裁剪/溢出。 Environment - Version: OpenClaw 2026.6.1 (2e08f0f) - UI: Control UI / WebChat (desktop)...
- Add Slack-safe progress visibility mode — Summary Slack progress/native task-card rendering needs a Slack-safe visibility mode so shared Slack threads do not expose raw tool calls, command text, local paths, or other operational details by...
- [AI] fix(memory): eliminate absent-window race in index swap + prevent auto-create empty DB — Summary If the running gateway opens main.sqlite while that file is briefly absent during the non-atomic index swap (or before any index exists), new DatabaseSync(path) auto-creates an empty database...
- [Bug]: Windows 11 - Claude CLI provider fails with spawn claude ENOENT — Bug type Regression (worked before, now fails) Beta release blocker No Summary OpenClaw fails to execute Claude Code on Windows 11 with "spawn claude ENOENT" even though Claude Code is installed and...
- fix(cli-runner): scope claude-cli queue to live-session owner identity (#91946) — Summary Fixes #91946. Replaces the workspace-scoped queue key that fresh claude-cli runs share with the same owner identity that claude-live-session.ts already uses for its live-session map....
- feat: add tool search directory mode — Summary - Add tools.toolSearch.mode: "directory" so large catalogs can expose a compact prompt directory plus tool_describe/tool_call controls while deferring most schemas. - Hydrate likely or...
- Feat/acp hub delegated sessions — Summary - Problem: Hub sessions (WebChat, WeChat, and other main chat surfaces) need persistent external ACP harness workers without binding a Discord/Telegram thread. The existing ACP persistent...
- [Bug]: After updating to OpenClaw 2026.6.1, legacy state migration warnings keep appearing even after running `openclaw doctor --fix`. — Bug type Regression (worked before, now fails) Beta release blocker No Summary After updating to OpenClaw 2026.6.1, legacy state migration warnings keep appearing even after running openclaw doctor...
- fix(ui): treat started managed-service handoff as pending update instead of skipped — Summary Fixes #87889. When a user clicks "Update now" in the Dashboard on a global install with a managed-service supervisor, the Gateway starts a managed-service handoff and returns ok=true with...
- fix(doctor): show per-step progress spinners during update — Summary - openclaw doctor offers a git update before running checks, but the update (fetch/rebase/build/ui:build/doctor) ran behind a single static note with zero progress indication for several...
- fix(telegram): avoid silent truncation in partial streaming finals — Summary Describe the problem and fix in 2–5 bullets: - Problem: Telegram partial-stream finals could be marked as delivered before continuation chunks were actually sent, which could suppress visible...
- [Feature]: logical chat history across rolled sessions — Summary Add a logical-history mode to chat.history so long-lived chats can retain continuity across daily, idle, or manual session rollovers. Problem to solve A stable session key such as...
- [codex] Avoid nested filesystem sandbox for OpenClaw sandbox tools — Summary When OpenClaw owns Codex native tool execution through the sandbox exec-server, start the Codex app-server thread with danger-full-access so Codex does not also apply its own filesystem...
- fix(cron): expose real SQLite database path in status output — Summary Since OpenClaw 2026.6.1, cron jobs are stored in the shared state SQLite database (~/.openclaw/state/openclaw.sqlite), but the openclaw cron status command still reports the legacy JSON...
- fix #69443: [Bug] Subagent RPC callback to WeChat session key routed to main session instead — Summary - Fixes #69443: subagent callback RPCs that pass a full agent-scoped session key in to should select that session instead of falling back to agent:main:main. - Promotes complete agent:...
- fix(plugins): rescan storm in "/models" call (regression shipped since v2026.5.18) — TL;DR: same box, same Telegram /model workflow as #90747, the other half of the stall: with the catalog forced into full discovery, every /models step re-ran the full plugin manifest scan several...
- fix(cron): expose real SQLite database path in status output — Summary Since OpenClaw 2026.6.1, cron jobs are stored in the shared state SQLite database (~/.openclaw/state/openclaw.sqlite), but the openclaw cron status command still reports the legacy JSON...
- Feature: Postcondition checks for cron jobs (deterministic verification of agent outcomes) — Problem Cron jobs with sessionTarget: "isolated" and payload.kind: "agentTurn" are inherently non-deterministic. The agent may stop prematurely, hallucinate completion, or silently underperform — and...
- [Feature]: Skill Permission Manifest Standard (skill.yaml) — Problem Skills currently run with full trust - there's no standard way for a skill to declare what permissions it needs, and no way for users to review permissions before installation. Recent...
- [Feature]: Markdown validation for prompt files — Summary Markdown files used for system prompts should follow rules to ensure they don't contain any patterns that could lead to prompt injections. While not everyone might agree on what is an...
- feat: add linux/riscv64 Docker image to multi-platform build — Component Docker release pipeline (docker-release.yml) Problem statement OpenClaw currently ships Docker images for linux/amd64 and linux/arm64. Users running RISC-V 64-bit hardware — developer...
- Feature request: session:reset hook event for memory persistence on auto-resets — Problem The session-memory bundled hook only fires on command:new events. When sessions auto-reset via: - Daily reset (session.reset.mode: "daily", atHour: 4) - Idle timeout...
- [Feature]: Memory/Context Improvements (metrics + global semantic search + conversation chaining + preload on restart) — Summary This issue groups several related improvements around agent memory, cross-session context, and observability. They came up while evolving a local-first personal assistant setup. Proposed...
- OpenClaw repo stats for 2026-06-11 — 378098 stars, 79063 forks, 2522 contributors
- OpenClaw sponsors (112 total) — 112 sponsors supporting OpenClaw
- fix: remove homepage suggestion chips (#2590)
- feat: rank skills by installs and dedupe install telemetry (#2535)
- fix: normalize ClawScan review visibility
- fix: reject malformed github oauth ids
- fix: prevent GitHub email account linking
- docs: remove stale skill folder guidance — Remove obsolete AGENTS.md guidance that treated repo skill additions as a standing contributor-policy issue.
- Add skill: brewpage-publish
- openclaw-cli 2026.6.5 on Homebrew — Your own personal AI assistant
- ykarout/Qwen3.5-9b-Opus-Openclaw-Distilled-GGUF — text-generation
- mradermacher/Qwen3.5-9b-Opus-Openclaw-Distilled-i1-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...
- OpenClaw is a security nightmare dressed up as a daydream
- OpenClaw isn't fooling me. I remember MS-DOS
- Nanobot: Ultra-Lightweight Alternative to OpenClaw
- Stop sending every AI coding request to the expensive model — AI coding tools are powerful. But they’re also wasteful. A tiny helper-function question does not...
- Self-hosting OpenClaw: a money trap and two silent failures — Three operational gotchas from self-hosting OpenClaw as a Signal press-review bot — an OpenRouter routing trap that spends your credits on a provider you never picked, a signal-cli crash on ARM64,...
- Telegram Integration - 0$ Personal Agentic AI Assistant - Part 5 — Introduction Part 5 of the Zero Dollar AI Assistant series - Your Personal AI Assistant...
- My Telegram bot stopped replying after OpenClaw 2026.6.1 — it was a full disk, not GPT-5 — A Telegram bot failure in OpenClaw 2026.6.1 looked like a GPT-5 issue but turned out to be ENOSPC: no space left on device, write. This post walks t
- My OpenClaw agent started writing nonsense and the real fix was a kill switch, not a better prompt — If your OpenClaw agent starts producing gibberish and /abort stops working, the fix is not a better prompt. The fix is architecture: disposable git
- I think I found the first real reason to build AI agent workflows in OpenClaw — A practical case for using OpenClaw to automate receipt-to-ledger workflows: bounded inputs and outputs, clear human review gates, and a better split
- Deploy OpenClaw on AWS Lightsail without manual headaches — This article is an English adaptation of the original Spanish post: Desplegar OpenClaw en AWS...
- How I Explored Hemers Agent — My First Impressions — Today, I spent some time working with Hemers Agent, and I wanted to share a few simple thoughts from my hands-on experience. I’ve been…Continue reading on Medium »
- AI Agents Are About to Get Their Own Internet — Meet the Agent Network for OpenClaw and Hermes — A decentralized, real-time, cross-framework AI Agent civilization running locally, collaborating globally — no cloud required.Continue reading on Medium »
- From Bedrock Posture to Agent Posture: Securing the Layer Between the Model and Your Systems — The moment the posture review misses half the systemContinue reading on Medium »
- OpenClaw’s AI Assistant Tricks Reveal User Data — What Happened?Continue reading on Medium »
- Navigating the 2026 Agent Ecosystem: LangGraph vs. OpenClaw vs. Google Agent SDK (ADK) — Choosing the right framework for deterministic orchestration, autonomous execution, or cloud-native ecosystem scaling.Continue reading on Medium »
- OCR for Notebooks: Why I Talk to an AI Instead — I use a physical notebook to collect ideas. Eventually I add these notes to Obsidian. The problem is that eventually may be a long time.Continue reading on Medium »
- The Ghost in the Machine: How Do We Catch an AI Criminal That Can Delete Itself? — Let me start with a question that no one in law enforcement wants to answer in public.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...
- 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...
- 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...
- RT @steipete: Part of the OpenClaw hardening work is reducing surface risk; for some media conversion we had to shell ou
- 1/ OpenClaw 2026.6.5 发布，核心是「防自我污染」： MCP 工具返回的 resource_link/resource/audio/异常 image 等富内容，现在会在 materialize 边界统一强制归一化，避免触 — 1/ OpenClaw 2026.6.5 发布，核心是「防自我污染」： MCP 工具返回的 resource_link/resource/audio/异常 image 等富内容，现在会在 materialize 边界统一强制归一化，避免触发 Anthropic 400 并污染会话历史。典型的边界收敛思路。
- Agent 生态日报 · 6/11 今天主线很清晰：可靠性与边界治理正在取代「能力炫技」，成为框架和工具竞争的真正战场。从 OpenClaw 的 MCP 结果归一化，到一批 Claude Code/Codex 「飞行记录仪」工具集中冒头， — Agent 生态日报 · 6/11 今天主线很清晰：可靠性与边界治理正在取代「能力炫技」，成为框架和工具竞争的真正战场。从 OpenClaw 的 MCP 结果归一化，到一批 Claude Code/Codex 「飞行记录仪」工具集中冒头，大家都在解同一个问题——让 Agent 在生产里不崩、不漏、可观测。🧵
- RT @anything: BREAKING: agents are taking over app dev Claude Code, OpenClaw, Codex, and Hermes can now design, build,
- @LiteLLM openclaw/openclaw 2026.6.6 shipped after sustained development. Session handling and MCP channel bridging fixes
- @steipete with something like openclaw where its humans finding and reporting problems/bugs your ai can do all the time,
- @nexusxrobotics @Conste11ation @openclaw the on-chain commit hash still shows the pre-edit state even if the local log c
- RT @heynavtoor: 10 GitHub repos that defined 2026 so far. Bookmark this list. 1. OpenClaw Peter Steinberger went from
- RT @kicoes: Para los que se montan un OpenClaw en el portátil viejo. Tenerlo encendido 24x7 te costará solo en electrici
- 终端跑 AI 代理（Claude Code / OpenClaw 等）很强，但配置繁琐、模型路由乱、权限不透明，难道只能忍受满屏的黑窗口？ 试试 freestylefly/wesight，一款将 CLI AI 代理直接变成可视化“桌面工作
- @MichaelGannotti @openclaw hope gabriel does not start asking for equity after this late night session
- @BuildWithxAI openclaw
- @devbjj Será que dá pra fazer no openclaw? Qual plano do Claude você pegou pra fazer isso? Tenho vontade de criar um ser
- AIエージェントもフィッシング詐欺に引っかかる？ 米セキュリティ企業がOpenClawで検証 結果は……(ITmedia NEWS) https://t.co/Ps1K0H4K37 — AIエージェントもフィッシング詐欺に引っかかる？ 米セキュリティ企業がOpenClawで検証 結果は……(ITmedia NEWS) https://t.co/Ps1K0H4K37
- RT @steipete: Part of the OpenClaw hardening work is reducing surface risk; for some media conversion we had to shell ou
- @david_bonilla @openclaw No sé hasta qué punto es taller pero me he quedado con muchas ganas de ver lo que contó ayer @d
- THIS DEVELOPER USED OPENCLAW AGENTS TO RUN HIS B2B BUSINESS VIA TELEGRAM AND MADE $15,000/MONTH he doesn't write prompt
- 🎬 OpenClaw Video Stack 一键部署：让 OpenClaw 拥有完整的 AI 视频生产能力 基于 OpenClaw + Seedance 2.0（字节跳动即梦）+ FFmpeg，从一篇文章自动生成 30 秒短视频（含分镜 — 🎬 OpenClaw Video Stack 一键部署：让 OpenClaw 拥有完整的 AI 视频生产能力 基于 OpenClaw + Seedance 2.0（字节跳动即梦）+ FFmpeg，从一篇文章自动生成 30 秒短视频（含分镜、AI 视频、中文字幕、配音、背景音乐）。 https://t.co/Bvmxq6HTzw
- https://t.co/E74Cy2SUW5
- OpenClaw AI agent leaked sensitive data despite strict security settings An AI agent with strict security settings hand
Trending on X
- RT @steipete: Part of the OpenClaw hardening work is reducing surface risk; for some media conversion we had to shell ou
- 1/ OpenClaw 2026.6.5 发布，核心是「防自我污染」： MCP 工具返回的 resource_link/resource/audio/异常 image 等富内容，现在会在 materialize 边界统一强制归一化，避免触 — 1/ OpenClaw 2026.6.5 发布，核心是「防自我污染」： MCP 工具返回的 resource_link/resource/audio/异常 image 等富内容，现在会在 materialize 边界统一强制归一化，避免触发 Anthropic 400 并污染会话历史。典型的边界收敛思路。
- Agent 生态日报 · 6/11 今天主线很清晰：可靠性与边界治理正在取代「能力炫技」，成为框架和工具竞争的真正战场。从 OpenClaw 的 MCP 结果归一化，到一批 Claude Code/Codex 「飞行记录仪」工具集中冒头， — Agent 生态日报 · 6/11 今天主线很清晰：可靠性与边界治理正在取代「能力炫技」，成为框架和工具竞争的真正战场。从 OpenClaw 的 MCP 结果归一化，到一批 Claude Code/Codex 「飞行记录仪」工具集中冒头，大家都在解同一个问题——让 Agent 在生产里不崩、不漏、可观测。🧵
- RT @anything: BREAKING: agents are taking over app dev Claude Code, OpenClaw, Codex, and Hermes can now design, build,
- @LiteLLM openclaw/openclaw 2026.6.6 shipped after sustained development. Session handling and MCP channel bridging fixes
- @steipete with something like openclaw where its humans finding and reporting problems/bugs your ai can do all the time,
- @nexusxrobotics @Conste11ation @openclaw the on-chain commit hash still shows the pre-edit state even if the local log c
- RT @heynavtoor: 10 GitHub repos that defined 2026 so far. Bookmark this list. 1. OpenClaw Peter Steinberger went from
- RT @kicoes: Para los que se montan un OpenClaw en el portátil viejo. Tenerlo encendido 24x7 te costará solo en electrici
- 终端跑 AI 代理（Claude Code / OpenClaw 等）很强，但配置繁琐、模型路由乱、权限不透明，难道只能忍受满屏的黑窗口？ 试试 freestylefly/wesight，一款将 CLI AI 代理直接变成可视化“桌面工作
- @MichaelGannotti @openclaw hope gabriel does not start asking for equity after this late night session
- @BuildWithxAI openclaw
- @devbjj Será que dá pra fazer no openclaw? Qual plano do Claude você pegou pra fazer isso? Tenho vontade de criar um ser
- AIエージェントもフィッシング詐欺に引っかかる？ 米セキュリティ企業がOpenClawで検証 結果は……(ITmedia NEWS) https://t.co/Ps1K0H4K37 — AIエージェントもフィッシング詐欺に引っかかる？ 米セキュリティ企業がOpenClawで検証 結果は……(ITmedia NEWS) https://t.co/Ps1K0H4K37
- RT @steipete: Part of the OpenClaw hardening work is reducing surface risk; for some media conversion we had to shell ou
- @david_bonilla @openclaw No sé hasta qué punto es taller pero me he quedado con muchas ganas de ver lo que contó ayer @d
- THIS DEVELOPER USED OPENCLAW AGENTS TO RUN HIS B2B BUSINESS VIA TELEGRAM AND MADE $15,000/MONTH he doesn't write prompt
- 🎬 OpenClaw Video Stack 一键部署：让 OpenClaw 拥有完整的 AI 视频生产能力 基于 OpenClaw + Seedance 2.0（字节跳动即梦）+ FFmpeg，从一篇文章自动生成 30 秒短视频（含分镜 — 🎬 OpenClaw Video Stack 一键部署：让 OpenClaw 拥有完整的 AI 视频生产能力 基于 OpenClaw + Seedance 2.0（字节跳动即梦）+ FFmpeg，从一篇文章自动生成 30 秒短视频（含分镜、AI 视频、中文字幕、配音、背景音乐）。 https://t.co/Bvmxq6HTzw
- https://t.co/E74Cy2SUW5
- OpenClaw AI agent leaked sensitive data despite strict security settings An AI agent with strict security settings hand
Releases
- OpenClaw 2026.6.6-beta.1 — 2026.6.6 Highlights - Security boundaries are substantially tighter across transcripts, sandbox binds, host environment inheritance, MCP stdio, Codex HTTP access, native search policy, elevated...
- OpenClaw repo stats for 2026-06-11 — 378098 stars, 79063 forks, 2522 contributors
- openclaw-cli 2026.6.5 on Homebrew — Your own personal AI assistant
Community
- #92109: [Bug]: EmbeddedAttemptSessionTakeoverError caused by Btrfs ctimeNs instability — Summary Fix the session-file fence so Btrfs ctimeNs-only metadata drift does not falsely trip EmbeddedAttemptSessionTakeoverError during cron/subagent prompt-lock release. The maintained branch keeps...
- [Bug]: Telegram isolated ingress deletes spooled updates when the turn fails — errors are swallowed so handleUpdate never rejects (silent message loss) — Summary In Telegram isolated polling ingress mode (default-on), a spooled inbound update is permanently deleted even when processing it failed, because the spool consumer keys deletion on...
- fix(agents): handle "Not supported model" error format in isModelNotFoundErrorMessage — Summary Fixes #92118 isModelNotFoundErrorMessage() in src/agents/live-model-errors.ts did not recognize the "Not supported model " error format returned by some OpenAI-compatible provider gateways...
- Bug: WebChat streaming display shows tool calls/results out of order — requires F5 refresh to correct — Environment - OpenClaw Gateway: 2026.6.5 - Browser: Chrome Problem In the OpenClaw WebChat UI, the real-time streaming display of a multi-step agent turn does not maintain correct ordering. A typical...
- isModelNotFoundErrorMessage does not match "Not supported model" error format — Description isModelNotFoundErrorMessage() in src/agents/live-model-errors.ts does not recognize the error format "Not supported model " returned by some OpenAI-compatible providers (e.g. internal...
- Flaky / failing: simple-completion-runtime > can preserve asynchronous provider model discovery (introduced 25ca39e876) — Flaky / failing test on main: simple-completion-runtime.test.ts > prepareSimpleCompletionModel > can preserve asynchronous provider model discovery Summary The test added in commit 25ca39e876...
- [Bug]: Collapsed sidebar + window resize breaks chat composer input area layout — Summary 折叠导航侧边栏（.shell--nav-collapsed）后左右拉伸/缩放浏览器窗口，WebChat 界面底部输入框（compose 区域）左侧的控制元素无法正确收缩，导致输入框左侧布局错位/裁剪/溢出。 Environment - Version: OpenClaw 2026.6.1 (2e08f0f) - UI: Control UI / WebChat (desktop)...
- Add Slack-safe progress visibility mode — Summary Slack progress/native task-card rendering needs a Slack-safe visibility mode so shared Slack threads do not expose raw tool calls, command text, local paths, or other operational details by...
- [AI] fix(memory): eliminate absent-window race in index swap + prevent auto-create empty DB — Summary If the running gateway opens main.sqlite while that file is briefly absent during the non-atomic index swap (or before any index exists), new DatabaseSync(path) auto-creates an empty database...
- [Bug]: Windows 11 - Claude CLI provider fails with spawn claude ENOENT — Bug type Regression (worked before, now fails) Beta release blocker No Summary OpenClaw fails to execute Claude Code on Windows 11 with "spawn claude ENOENT" even though Claude Code is installed and...
- fix(cli-runner): scope claude-cli queue to live-session owner identity (#91946) — Summary Fixes #91946. Replaces the workspace-scoped queue key that fresh claude-cli runs share with the same owner identity that claude-live-session.ts already uses for its live-session map....
- feat: add tool search directory mode — Summary - Add tools.toolSearch.mode: "directory" so large catalogs can expose a compact prompt directory plus tool_describe/tool_call controls while deferring most schemas. - Hydrate likely or...
- Feat/acp hub delegated sessions — Summary - Problem: Hub sessions (WebChat, WeChat, and other main chat surfaces) need persistent external ACP harness workers without binding a Discord/Telegram thread. The existing ACP persistent...
- [Bug]: After updating to OpenClaw 2026.6.1, legacy state migration warnings keep appearing even after running `openclaw doctor --fix`. — Bug type Regression (worked before, now fails) Beta release blocker No Summary After updating to OpenClaw 2026.6.1, legacy state migration warnings keep appearing even after running openclaw doctor...
- fix(ui): treat started managed-service handoff as pending update instead of skipped — Summary Fixes #87889. When a user clicks "Update now" in the Dashboard on a global install with a managed-service supervisor, the Gateway starts a managed-service handoff and returns ok=true with...
- fix(doctor): show per-step progress spinners during update — Summary - openclaw doctor offers a git update before running checks, but the update (fetch/rebase/build/ui:build/doctor) ran behind a single static note with zero progress indication for several...
- fix(telegram): avoid silent truncation in partial streaming finals — Summary Describe the problem and fix in 2–5 bullets: - Problem: Telegram partial-stream finals could be marked as delivered before continuation chunks were actually sent, which could suppress visible...
- [Feature]: logical chat history across rolled sessions — Summary Add a logical-history mode to chat.history so long-lived chats can retain continuity across daily, idle, or manual session rollovers. Problem to solve A stable session key such as...
- [codex] Avoid nested filesystem sandbox for OpenClaw sandbox tools — Summary When OpenClaw owns Codex native tool execution through the sandbox exec-server, start the Codex app-server thread with danger-full-access so Codex does not also apply its own filesystem...
- fix(cron): expose real SQLite database path in status output — Summary Since OpenClaw 2026.6.1, cron jobs are stored in the shared state SQLite database (~/.openclaw/state/openclaw.sqlite), but the openclaw cron status command still reports the legacy JSON...
- fix #69443: [Bug] Subagent RPC callback to WeChat session key routed to main session instead — Summary - Fixes #69443: subagent callback RPCs that pass a full agent-scoped session key in to should select that session instead of falling back to agent:main:main. - Promotes complete agent:...
- fix(plugins): rescan storm in "/models" call (regression shipped since v2026.5.18) — TL;DR: same box, same Telegram /model workflow as #90747, the other half of the stall: with the catalog forced into full discovery, every /models step re-ran the full plugin manifest scan several...
- fix(cron): expose real SQLite database path in status output — Summary Since OpenClaw 2026.6.1, cron jobs are stored in the shared state SQLite database (~/.openclaw/state/openclaw.sqlite), but the openclaw cron status command still reports the legacy JSON...
- Feature: Postcondition checks for cron jobs (deterministic verification of agent outcomes) — Problem Cron jobs with sessionTarget: "isolated" and payload.kind: "agentTurn" are inherently non-deterministic. The agent may stop prematurely, hallucinate completion, or silently underperform — and...
- [Feature]: Skill Permission Manifest Standard (skill.yaml) — Problem Skills currently run with full trust - there's no standard way for a skill to declare what permissions it needs, and no way for users to review permissions before installation. Recent...
- [Feature]: Markdown validation for prompt files — Summary Markdown files used for system prompts should follow rules to ensure they don't contain any patterns that could lead to prompt injections. While not everyone might agree on what is an...
- feat: add linux/riscv64 Docker image to multi-platform build — Component Docker release pipeline (docker-release.yml) Problem statement OpenClaw currently ships Docker images for linux/amd64 and linux/arm64. Users running RISC-V 64-bit hardware — developer...
- Feature request: session:reset hook event for memory persistence on auto-resets — Problem The session-memory bundled hook only fires on command:new events. When sessions auto-reset via: - Daily reset (session.reset.mode: "daily", atHour: 4) - Idle timeout...
- [Feature]: Memory/Context Improvements (metrics + global semantic search + conversation chaining + preload on restart) — Summary This issue groups several related improvements around agent memory, cross-session context, and observability. They came up while evolving a local-first personal assistant setup. Proposed...
- OpenClaw sponsors (112 total) — 112 sponsors supporting OpenClaw
- fix: remove homepage suggestion chips (#2590)
- feat: rank skills by installs and dedupe install telemetry (#2535)
- fix: normalize ClawScan review visibility
- fix: reject malformed github oauth ids
- fix: prevent GitHub email account linking
- docs: remove stale skill folder guidance — Remove obsolete AGENTS.md guidance that treated repo skill additions as a standing contributor-policy issue.
- Add skill: brewpage-publish
News
- ykarout/Qwen3.5-9b-Opus-Openclaw-Distilled-GGUF — text-generation
- mradermacher/Qwen3.5-9b-Opus-Openclaw-Distilled-i1-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...
- OpenClaw is a security nightmare dressed up as a daydream
- OpenClaw isn't fooling me. I remember MS-DOS
- Nanobot: Ultra-Lightweight Alternative to OpenClaw
- Stop sending every AI coding request to the expensive model — AI coding tools are powerful. But they’re also wasteful. A tiny helper-function question does not...
- Self-hosting OpenClaw: a money trap and two silent failures — Three operational gotchas from self-hosting OpenClaw as a Signal press-review bot — an OpenRouter routing trap that spends your credits on a provider you never picked, a signal-cli crash on ARM64,...
- Telegram Integration - 0$ Personal Agentic AI Assistant - Part 5 — Introduction Part 5 of the Zero Dollar AI Assistant series - Your Personal AI Assistant...
- My Telegram bot stopped replying after OpenClaw 2026.6.1 — it was a full disk, not GPT-5 — A Telegram bot failure in OpenClaw 2026.6.1 looked like a GPT-5 issue but turned out to be ENOSPC: no space left on device, write. This post walks t
- My OpenClaw agent started writing nonsense and the real fix was a kill switch, not a better prompt — If your OpenClaw agent starts producing gibberish and /abort stops working, the fix is not a better prompt. The fix is architecture: disposable git
- I think I found the first real reason to build AI agent workflows in OpenClaw — A practical case for using OpenClaw to automate receipt-to-ledger workflows: bounded inputs and outputs, clear human review gates, and a better split
- Deploy OpenClaw on AWS Lightsail without manual headaches — This article is an English adaptation of the original Spanish post: Desplegar OpenClaw en AWS...
- How I Explored Hemers Agent — My First Impressions — Today, I spent some time working with Hemers Agent, and I wanted to share a few simple thoughts from my hands-on experience. I’ve been…Continue reading on Medium »
- AI Agents Are About to Get Their Own Internet — Meet the Agent Network for OpenClaw and Hermes — A decentralized, real-time, cross-framework AI Agent civilization running locally, collaborating globally — no cloud required.Continue reading on Medium »
- From Bedrock Posture to Agent Posture: Securing the Layer Between the Model and Your Systems — The moment the posture review misses half the systemContinue reading on Medium »
- OpenClaw’s AI Assistant Tricks Reveal User Data — What Happened?Continue reading on Medium »
- Navigating the 2026 Agent Ecosystem: LangGraph vs. OpenClaw vs. Google Agent SDK (ADK) — Choosing the right framework for deterministic orchestration, autonomous execution, or cloud-native ecosystem scaling.Continue reading on Medium »
- OCR for Notebooks: Why I Talk to an AI Instead — I use a physical notebook to collect ideas. Eventually I add these notes to Obsidian. The problem is that eventually may be a long time.Continue reading on Medium »
- The Ghost in the Machine: How Do We Catch an AI Criminal That Can Delete Itself? — Let me start with a question that no one in law enforcement wants to answer in public.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...
- 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...
Security
- 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...
Don't miss what's next. Subscribe to Openclaw Newsletter: