documentation intermediate active

OpenClaw AGENTS.md

You build multi-agent OpenClaw pipelines on guesswork. Read the official spec first.

What breaks without openclaw agents documentation

Undocumented agent lifecycle. Missing event filters. Unhandled rejections crashing apps.

Correct agent patterns × official specification ÷ 15-minute read ÷ no architecture guessing = stable multi-agent pipelines.

openclaw agents documentation — what it actually does

01
Defines the OpenClaw agent model: stateful, event-driven, with lifecycle hooks.
02
Documents how to register agents with proper event filters.
03
Covers multi-agent orchestration patterns within a single application.
04
Guides AI coding tools on conventions when editing agent code.
05
Explains agent error handling to prevent unhandled rejection crashes.

Security check — openclaw agents documentation

Privacy score: 7/10 — accesses connected platform APIs only. Lock it: review OAuth scopes before install, confirm OpenClaw ≥1.3 for multi-agent features compatibility.

Quick start — openclaw agents documentation in 15–30 minutes to read and understand

Setup time: 15–30 minutes to read and understand

!
You need: Familiarity with OpenClaw core architecture

Install the package:

# Documentation file — no install required
1
Read the Agent Definition section
2
Understand the event subscription model
3
Review the agent registration API
4
Study the orchestration patterns
5
Implement a simple agent following the pattern
6
Test with openclaw dev --agent myagent

Troubleshooting openclaw agents documentation

1
1. Confusing OpenClaw agents with external AI agent frameworks — different abstractions
2
2. Registering agents without proper event filters causes performance issues
3
3. Not handling agent errors — unhandled rejections crash the entire app

Compatibility & status

Works with: OpenClaw ≥1.3 for multi-agent features intermediate Last updated: Oct 2025 MIT

Official docs →

View on GitHub →

FAQ — openclaw agents documentation

Are OpenClaw 'agents' the same as LLM agents?

No. OpenClaw agents are stateful event handlers. LLM inference is optional.

Can agents communicate with each other?

Yes, via the internal event bus using ctx.broadcast().

Where do I register a new agent?

In openclaw.config.js under the agents array.

Related — more like openclaw agents documentation

More by openclaw

Building agents without reading AGENTS.md causes silent performance bugs.

Read it before your multi-agent PR.

Get it on GitHub →