Deep-dive technical documentation for building production-grade autonomous infrastructure.
New to self-hosted AI? Follow our curated path to go from zero to a fully integrated WhatsApp/Slack assistant in one afternoon.
Debug OpenAI-compatible relays, local-model servers, Ollama `/v1`, vLLM, LiteLLM, tools, reasoning, and streaming mismatches with our new compatibility track.
Download a conservative openclaw.json preset for local, coding, or remote dashboard access. Start with a small blast radius, then verify with a minimal probe loop before you add channels, skills, or automation.
A task-first map of OpenClaw commands, plus the recurring questions from Reddit: gateway disconnected (4008), dashboard unauthorized, pairing required (1008), TUI loops, and “what do I run to debug this?”
A field guide to running OpenClaw on native Windows without guesswork. Learn why shell PATH and Scheduled Task PATH diverge, when `tools.exec.pathPrepend` helps or hurts, why `openclaw node run` can stall after printing PATH, and when WSL2 is the lower-risk choice.
A practical decision guide for Windows users: when native Windows is good enough, when WSL2 is the safer default, which install path to choose, and how to migrate without turning your setup into a mystery.
A practical step-by-step guide to fix OpenClaw when it suddenly stops using tools after recent updates. Learn how to check `tools.profile`, restore coding tools safely, and verify the agent can act again.
A layered checklist for the most common 'can't install' / 'command not found' / 'service won't start' failures. Covers Node version, global install permissions, PATH issues, WSL2 systemd, and Docker setup gotchas.
A high-signal checklist for when OpenClaw stops replying (TUI shows '(no output)', channels go quiet, or logs show 401/403/429). Covers config precedence, provider auth, model allowlists, relay API-mode mismatch, and rate-limit/billing traps.
A practical hardening guide for OpenClaw operators: reduce skill malware risk, defend against prompt injection from emails/web pages, and design safe-by-default workflows (separate accounts, least privilege, allowlists, audit trails).
Download a ready-made TASKS.md file for your OpenClaw workspace to track multi-step work, make cron runs leave evidence, and reduce 'it promised but I can't see what happened' failures.
Run a containerized OpenClaw Gateway with Docker Compose for an isolated, reproducible setup.
A practical guide to wiring OpenClaw to Telegram (Bot API) with safe defaults, plus the lessons that usually bite first-time users.
A practical operator checklist to control blast radius and token spend: start with a bounded tool profile, tighten DM policies, make runs leave evidence, and keep a kill switch. Designed for real life, not perfect autonomy.
A practical, step-by-step playbook for using the OpenClaw Control UI from another device (laptop/phone) without token mismatch, pairing loops (1008), or brittle public exposure. Pick one access pattern and make it stable.
A scannable triage flow for non-hardcore operators: map what you see (unauthorized, pairing required 1008, gateway disconnected 4008, no output, only chats) to the correct fix page, with a minimal command pack to prove each layer.
A practical guide to OpenClaw's built-in backup commands: choose the right backup scope, verify archives before risky changes, and keep a real rollback path for upgrades, migrations, and config surgery.
A practical guide to Brave's `llm-context` mode in OpenClaw: what it returns, when to use it instead of normal web search, how to think about freshness and region, and how to avoid common key and environment mistakes.
A practical guide to OpenClaw exec approvals: understand the difference between tool policy, elevated mode, allowlists, and safe bins; fix `allowlist miss`; and choose the right trust model for gateway and node hosts.
A practical guide to stable remote browser control in OpenClaw: run the gateway remotely, keep Chrome local, route through a node host, pin browser-capable nodes, and avoid the most common relay/probe confusions.
A practical guide to OpenClaw Talk Mode's silence timeout: understand the platform defaults, decide when to shorten or lengthen the pause window, and tune for dictation, natural conversation, or slow-thinking workflows.
A decision guide for choosing the right local or proxy AI API path for OpenClaw: native Ollama, Ollama /v1, llama.cpp, vLLM, LiteLLM, and generic OpenAI-compatible relays.
A practical compatibility matrix for using OpenClaw with self-hosted and proxy AI APIs: native Ollama, Ollama /v1, llama.cpp, vLLM, LiteLLM, and OpenAI-compatible relays.
A practical guide to using ClawHub as a skill discovery and distribution channel: how to evaluate maintainers, install in a low-risk way, upgrade without surprises, and keep a real rollback path when a skill fails or disappears.
A practical guide to using OpenClaw with Home Assistant without over-automating your house: where the boundary should live, what to delegate to an agent, and which risks to control first.
A practical guide to running multiple OpenClaw agents behind one gateway without turning your setup into guesswork. Learn how bindings, channel accounts, routing boundaries, workspace design, and Telegram-specific pitfalls fit together.
A practical guide to using Raspberry Pi as an OpenClaw edge node: which workloads fit, which do not, and how Pi compares with a NAS, Mac mini, and low-power mini PC.
Understand what Control UI pairing actually proves, why local and remote access behave differently, where users confuse pairing with auth, and how to troubleshoot trust failures without guessing.
Understand why OpenClaw recovery can keep looping after a successful OAuth renewal, and how to resync the live auth file, the provisioned copy, and safe-mode markers.
A practical guide to Control UI failures: why 'unauthorized' happens, how tokenized dashboard links work, what 'disconnected (1008): pairing required' really means, and how to connect safely to a remote gateway.
A pragmatic risk guide: why some integrations trigger 403/disabled-account events, how to reduce blast radius with separate accounts and least privilege, and how to design workflows that don’t depend on fragile consumer OAuth.
Stabilize OpenClaw browser runs (Playwright) when they time out: verify the browser service is alive, increase timeouts where appropriate, reduce page complexity, and fix common headless/container dependencies.
A pragmatic guide to OpenClaw automation reliability: how cron runs are scheduled and delivered, why jobs can 'advance nextRunAtMs' without executing, and how to design runs that always leave evidence and alert correctly.
Email is powerful and risky. This guide covers how to set up OpenClaw email integrations, why Gmail OAuth often needs re-auth, how to reduce blast radius (separate accounts/scopes), and when to use IMAP/SMTP or a dedicated mailbox provider instead.
A practical operability guide: how to capture high-signal logs, design runs that always leave evidence, and add a lightweight Kanban-style task board in your workspace so multi-step automations don’t feel invisible.
Fix 'memory not written', 'it said it saved files but nothing exists', and post-restart amnesia by understanding the state directory, workspace paths, container UID/GID, and how to make persistence reproducible across Docker/VPS/PaaS.
A practical integration guide for using OpenClaw with OpenAI/Anthropic-compatible relays and API proxies (NewAPI, OneAPI, AnyRouter, LiteLLM, vLLM): choose the right API mode, set baseUrl correctly, avoid config precedence traps, and debug 403/404/blank-output failures fast.
A platform-aware guide for Alibaba Cloud, Volcengine, Zeabur, Unraid, AWS, and Google Cloud: why image resets force token re-entry, how to persist OpenClaw state correctly, and how to debug STARTING / startup probe failed step-by-step.
A beginner-friendly but thorough guide to OpenClaw configuration: where openclaw.json lives, safe defaults, model/provider setup, gateway auth/networking, channels, plugins, and the most common config pitfalls.
A practical, opinionated checklist for new OpenClaw users: secure defaults, dashboard access, backups, updates, and the most common early gotchas.
A battle-tested update flow: snapshot state, update with checks, restart cleanly, and migrate to a new machine without losing credentials or sessions.
Connect a WhatsApp account to OpenClaw and restrict who can message your agent using allowlists and pairing.
Turn on optional plugins and explicitly allowlist tools so your agent can safely call them.
Our plugin SDK is open source. Contribute your own integrations or request a dedicated build from the core team.