Skip to main content
Actions are the capabilities available to your AI agents. Every tool call an agent makes — reading a file, sending an email, searching the web — is an action. The Action Library is where you configure which actions are available, how they’re authenticated, and what permissions they require.

​
The Three Layers

OpenSwarm organizes actions into three layers:

Built-in Actions

The core actions enabling a general agent — file operations, search,  commands, browser automation, and more.

Integrations

First-party connectors for services like Google Workspace, Twitter/X, and Reddit. These connect via OAuth or credentials.

MCP Actions

Any MCP (Model Context Protocol) server — installed from the public registry or added manually. This is the primary extension point.

​
How Actions Flow Into Agent Sessions

When an agent session starts, OpenSwarm assembles the full set of available actions:
  1. Core actions that are enabled are loaded directly into the session.
  2. All other actions (like web search, Reddit, Twitter/X, etc.) are loaded on-demand when the agent looks for more specialized tools.
Each action is also subject to its permission policy, which determines whether the agent can execute it freely, must ask the user first, or is blocked entirely.

​
What’s Next

Built-in Actions & Integrations

Learn about every built-in tool and first-party integration.

Permissions

Control what agents can do with granular permission policies.

MCP Registry

Browse, install, and configure MCP servers.