Documentation

Docs for the online OpenClaw tool, written for fast onboarding.

The OpenClawAi.run documentation focuses on clarity: how to start an instance, connect your model key, understand lifecycle rules, and troubleshoot issues quickly. If you are new to OpenClaw, start with the quick start steps below and then explore the deeper sections when you need them.

Search the docs (coming soon)

A built-in search experience will help you find instance lifecycle rules, BYOK setup steps, and troubleshooting answers. For now, use the section map below to jump to the most important guidance.

Quick start in three steps

The online OpenClaw tool is built for quick onboarding. These steps describe the shortest path from zero to a running instance.

1. Create your instance From the Start page, create an instance and wait for the status badge to show ready.
2. Add BYOK credentials Insert your OpenAI or Anthropic key in the console. Keys stay inside your instance and can be removed.
3. Launch a session Start a session from the console, run your first workflow, and watch logs update in real time.

If your instance takes longer than expected to boot, check the Status page for ongoing incidents or rate limits. The console will also surface a clear error message when capacity is full.

Doc sections you will use most

These are the core topics that matter for OpenClawAi.run users. Each card summarizes the decisions you need to make before running real workloads.

Getting started

What the hosted OpenClaw instance is, who it is for, and how to launch your first session.

Instances & sessions

Definitions, lifecycle states, and how to pause, restart, or delete instances safely.

Limits & fairness

Free beta quotas, runtime limits, and what happens when you hit capacity.

BYOK credentials

Where keys live, how to remove them, and how to minimize the risk of exposing secrets.

Security model

Isolation approach, access controls, and how we keep instances separated.

Troubleshooting

Common issues like slow starts, console timeouts, and how to verify instance health.

Instance lifecycle explained

Understanding lifecycle rules prevents surprise pauses or data loss. The hosted OpenClaw instance follows three simple rules: each instance is isolated, idle time triggers auto-pause, and deletion is permanent.

States you will see

Instances move through ready, running, paused, and deleted states. The console shows a clear badge and timestamp so you know whether a session is active or waiting. If a run takes longer than expected, check the logs for queueing or capacity warnings.

Idle policy

Idle instances auto-pause to keep free usage fair. After a month of inactivity, the instance is removed and data cannot be recovered. If you need to keep data longer, export results before deleting the instance.

These rules are designed to keep the platform safe for everyone while giving you predictable control over your environment. If you need a different lifecycle for business use, contact support to discuss options.

BYOK credentials and data handling

OpenClawAi.run is BYOK, meaning you bring the model key. We store credentials inside your instance only. You can remove them through the console or by deleting the instance entirely.

Where keys live

Credentials are stored inside the instance environment, not shared across users. This keeps control with you and reduces the risk of leakage outside your runtime.

How to remove keys

Delete the credentials through the console or delete the instance to clear everything at once. Once deleted, keys cannot be recovered.

For security details, including access control and isolation, read the Security & Privacy page.

Limits, quotas, and fair use

The free beta has explicit limits: one concurrent instance, 30 minutes maximum runtime per session, and auto-pause for idle instances. If you hit a limit, you will see a clear warning in the console. These rules keep the online OpenClaw tool stable while we scale.

Need more capacity? The docs will outline upgrade paths as they become available. For now, the best practice is to keep sessions short and delete instances when you are done.

Troubleshooting basics

Most issues fall into a few predictable categories. Use this list to diagnose quickly before contacting support.

Instance stuck on starting

Check the Status page, then retry. If the issue persists, record the timestamp and browser details.

Console not loading

Refresh the session and confirm your network allows WebSocket connections. Clear cache if needed.

Unexpected pause

Idle instances auto-pause. Restart the instance from the console if you are still working.

Model key errors

Verify the key is active and has sufficient quota. Remove and re-add it if needed.

If you cannot resolve an issue, contact support at support@openclawai.run with your instance ID, timestamp, and a short description of what you were doing.

Start with docs, then launch your instance

The documentation is built to be practical: short steps, clear limits, and transparent lifecycle rules. When the hosted OpenClaw instance opens, you will know exactly what to expect.