{AF} AgentForge OpenClaw setup, guides, and ready agents
Install guide Book setup

Tools and plugins

How to connect tools and plugins in OpenClaw without turning setup into chaos

In OpenClaw, plugins are not cosmetic. They can add tool definitions, CLI commands, HTTP routes, services, and extra extensions. The real point is not installing many of them, but knowing when they help and how not to break the gateway.

The practical difference

Core tools already ship with OpenClaw. Plugins extend the platform and can introduce commands, agent tools, integrations, and background services. In the official docs, the first command worth using is:

openclaw plugins list

The cleanest path

  1. Check what is already loaded with openclaw plugins list.
  2. Install one official plugin only if it serves the first workflow.
  3. Restart the gateway and configure the plugin in its own entry.
openclaw plugins install @openclaw/voice-call
openclaw plugins enable voice-call
openclaw plugins doctor

An important security detail

The OpenClaw docs are very clear: native plugins run in-process with the gateway and should be treated as trusted code. That means a bad plugin can destabilize the entire system.

  • install only plugins you understand or trust
  • do not add plugins before you have one clear working flow
  • use plugins.allow or a minimal selection when the setup gets more serious

When plugins actually help

They help when the use case requires a capability that is not already in the core, such as a specific integration or optional channel. They do not help just because you want “more AI”. The right order stays the same: first OpenClaw, then models, then one agent, then only the plugins that really matter.