- fix(claude-cli): apply thinking and keep live sessions warm — Summary High Level TLDR Claude CLI now advertises and applies thinking levels end to end. Captured live sessions keep the same Claude process warm for compatible turns so prompt-cache reuse survives;...
- fix(control-ui): config form save corrupts 64-bit id strings in string|number fields — Closes #126401 What Problem This Solves The schema-driven Control UI settings form could silently change stored 64-bit identifier strings into rounded JavaScript numbers when any unrelated form field...
- feat(perplexity): add search context size — What Problem This Solves Fixes #84872. The managed Perplexity web_search provider did not expose the low / medium / high search_context_size content-budget knob, so agents could not request a smaller...
- Aborted run leaves unpersisted in-memory transcript entry; all subsequent runs fail with 'Session transcript parent entry was not persisted'
- fix(agents): fall back from tool search code under Electron — Summary - detect Electron-hosted runtimes before selecting tool-search code mode - use the existing structured-tools fallback under Electron - preserve code mode for ordinary Node runtimes that...
- Gateway restart races with Windows Task Scheduler (MultipleInstancesPolicy=IgnoreNew), causes ECONNREFUSED — Environment - OS: Windows 10/11, Gateway installed as a Windows Scheduled Task (default on Windows) - OpenClaw version: 2026.7.1-2 - Node: v24.19.0 Steps to reproduce 1. Install OpenClaw via npm i -g...
- build(deps): bump the actions group across 1 directory with 10 updates — [//]: (dependabot-start) ⚠️ Dependabot is rebasing this PR ⚠️ Rebasing might not happen immediately, so don't worry if this takes some time. Note: if you make any changes to this PR yourself, they...
- fix(release): accept trusted Codex scan layouts — What Problem This Solves Resolves a release qualification failure where the trusted publishable-plugin security inventory rejected the frozen beta candidate's reviewed Codex execution sites after...
- CLI subprocess processes orphaned on gateway restart — in-memory supervisor registry lost without killing child processes — Title CLI subprocess processes orphaned on gateway restart — in-memory supervisor registry lost without killing child processes Summary When the OpenClaw gateway restarts (planned or crash), CLI...
- [Bug]: Standalone codex CLI login and gateway 'openai' profile are separate auth stores, and there is no headless way to refresh the gateway one — Environment - OpenClaw 2026.7.1-2 (0790d9f), Ubuntu 24.04 (headless VPS), Telegram channel, agent main, model claude-cli/claude-sonnet-5 - Standalone codex CLI installed via curl -fsSL...
- [Feature]: Allow hiding progress activity receipt while keeping commentary — Summary Add a per-channel option to hide the finalized progress activity receipt (for example, 💬 2 notes · 🛠️ 6 tool calls · ⏱️ 131s) while keeping live progress drafts and assistant-authored...
- fix(compaction): ignore decimal identifier fragments — What Problem This Solves Strict safeguard compaction could classify long digit runs inside decimal and scientific numeric values as exact identifiers. For example, metric=0.123456789 could add...
- fix(models): classify catalog timeouts by identity — What Problem This Solves Provider discovery failures were classified as helper timeouts by matching the error message text. A provider could therefore reject with similar wording and OpenClaw would...
- gateway: add a suppressed marker to sessions.changed when a silent-reply broadcast is dropped — Summary When an agent emits the silent-reply token (SILENT_REPLY_TOKEN = "NO_REPLY", per plugin-sdk/src/auto-reply/tokens.d.ts), the gateway's broadcast pipeline drops the session.message event....
- [Bug]: git updater: recreated upstream tags make `git fetch --tags` fail forever while `openclaw status` reports "up to date" from stale refs — Bug type Behavior bug (incorrect output/state without crash) Beta release blocker No Summary On a git-main install, after upstream recreated two existing tags (v2026.4.15-beta.1, v2026.5.3-beta.1),...
- [Feature]: Expose Swagger open api schema — Summary It would be nice to have open api schema to interact with open claw Problem to solve I would like to generate custom api client to interact with open claw programmatically Proposed solution...
- [Reliability]: Enforce atomic intent→execution flow to prevent orphan pre-action notices — Summary Guarantee that announced actions always end in explicit commit/abort status. Problem to solve Pre-action notices can be emitted without guaranteed execution under interruptions. This creates...
- Auto-migration silently switches default model to a paid `-preview` variant on upgrade, causing unintended billing — Summary On upgrade from 2026.3.2 → 2026.3.22, my ~/.openclaw/openclaw.json was silently mutated such that agents.defaults.model.primary changed from google/gemini-2.5-pro to...
- [Bug]: Subagents can appear completed before the underlying delegated work is actually finished — Bug type Behavior bug (incorrect output/state without crash) Summary OpenClaw subagents can appear finished in the UI/session flow before the underlying delegated work is actually complete, which...
- Context engine treats terminal blocked outbox rows as pending work — Summary An unrecoverable context-engine turn advancement is retained in context_engine_turn_outbox with payload_json.state = "blocked". That row is useful terminal audit evidence, but the...
- fix(opencode): auto-surface live models via shared projector — Endpoint already returns new ids but catalog dropped them: - Zen: projectOpencodeZenLiveModels filtered to staticModels.get(id) only → gemini-3.7-flash, grok-4.6, hy3-free, muse-spark-1.2 dropped...
- contextPruning (mode: cache-ttl) never fires on OpenAI models: isCacheTtlEligibleProvider excludes OpenAI, so the idle-gap tool-result firebreak is dead for the highest-volume provider — Summary The contextPruning mode: "cache-ttl" firebreak — which prunes stale tool-results after an idle gap so post-idle requests don't re-cache oversized history — never executes for OpenAI models....
- Feature Request: recursive subagent listing (--depth / --recursive) — Problem With maxSpawnDepth: 2, we use the orchestrator pattern: main agent -> project manager (depth-1) -> worker experts (depth-2). The main agent needs visibility into depth-2 workers to monitor...
- Feature request: support Anthropic's mid-conversation tool-changes beta for cache-preserving tool swaps — Summary Anthropic shipped a beta (live now, ~2 weeks as of this writing) called mid-conversation tool changes, available on Claude Opus 5, Opus 4.8, Fable 5, and Mythos 5. It lets you add/remove...
- fix(imessage): reject ambiguous outbound recipients — What Problem This Solves Fixes an issue where an unqualified iMessage recipient containing both letters and digits could be reinterpreted as a phone number after its non-digit characters were...
- fix(diffs): guard malformed tool arguments — Summary - Guard the production diffs tool against null, undefined, array, and primitive raw arguments by reusing the existing non-array record coercion helper. - Preserve existing diff rendering,...
- OpenClaw repo stats for 2026-08-20 — 386842 stars, 81264 forks, 3112 contributors
- OpenClaw sponsors (96 total) — 96 sponsors supporting OpenClaw
- fix: accept sourced repo for skill eval claims — Resolve GitHub source repo through githubSkillSources during SkillEvaluator claim stale checks, and let the temporary doca prioritizer reset its stale-version row.
- fix: abort hung OG metadata fetches and archive metrics (#3471)
- docs: update Git & GitHub skill count to 160
- Add skill: programmingwtf/repo-standardizer
- openclaw-cli 2026.7.1 on Homebrew — Your own personal AI assistant
- 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...
- OpenClaw is what Apple intelligence should have been
- OpenClaw surpasses React to become the most-starred software project on GitHub
- Complete OpenClaw Installation Guide with Telegram Communication Channel — In 2026, agentic AI tools and frameworks generated much interest from technical and non-technical...
- Saving 50-70% on Coding Agent Bills: Protecting Prompt Cache Affinity with VMR — TL;DR — With per-token pricing now the default everywhere, the biggest line item in a coding-agent...
- Lessons from a 2,734-tick earning loop on OpenClaw — I let an AI agent run on my Mac mini for two months. It woke up every five minutes, picked one...
- Setting Up an OpenClaw Agent o… — Originally published at norvik.tech Introduction Explore the setup of an OpenClaw agent...
- How I detected and patched 12 autonomous-agent failure modes — Field data After running 33 autonomous agents in production, the failure modes cluster....
- 12 defensive patterns I shipped after OpenClaw 2026.4.14 broke my agents — What broke OpenClaw 2026.4.14 took out lossless-claw (#66591) and active-memory (#66849)....
- Your Agent Queue Is Not a Workflow Until It Tracks Execution State — An AI agent can receive the same job twice even when your queue is behaving correctly. A worker may...
- I thought OpenClaw would just install and instead I got dropped into provider hell — A practical DEV post on why OpenClaw installs often fail at the provider layer, not the package layer. The piece argues for a boring but reliable setu
- Your OpenClaw Deployment Is Not Portable Until You Rebuild It — An OpenClaw deployment can be healthy and still be impossible to recover. The dangerous assumption...
- 5 Best OpenClaw VPS Provider in India — Introduction OpenClaw is changing how people use AI assistants. Instead of opening a...
- I thought running OpenClaw from my phone was the goal, then I realized the phone should only say yes or no — A practical argument for using your phone as an approval layer for OpenClaw and agent workflows, not as the full control plane. Covers architecture, o
- OpenClaw vs Subscription AI Assistants: Real Token Costs, Capability Gaps and the Usage Level Where Self-Hosting Wins — OpenClaw wins on cost only at the two extremes of usage: very light use, where a metered token bill...
- The Status Quo of AI in Software Development (2026) — Artificial Intelligence in 2026: From Companion to Infrastructure Artificial Intelligence...
- Complete OpenClaw Installation Guide with Telegram Communication Channel — In 2026, agentic AI tools and frameworks generated much interest from technical and non-technical users alike. On the forefront of all of…Continue reading on Medium »
- [IQuest / RUC] — 블랙박스 에이전트 하네스(Harness) 상에서의 강화학습: ClawGym II 프레임워크 해부 — 1. 한눈에 보는 핵심 요약 (Executive Summary)Continue reading on Medium »
- Mastering OpenClaw for AI Agent Automation — Course titleContinue reading on Medium »
- Kimi — AI assistant supports complex math, coding, image understanding, and reasoning with up to 128k token processing, short and long chain-of-thought tasks, visual and text input, accurate benchmarks,...
- 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...