> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agent-loadout.com/llms.txt
> Use this file to discover all available pages before exploring further.

# MCP Server — Connect Any AI Client to Agent Loadout

> Connect Claude Code, Codex, Cursor, ChatGPT, or any MCP client to Agent Loadout's hosted MCP server. No local installation required.

Agent Loadout exposes a hosted MCP server that any compatible AI client can connect to. Point your client at the server URL, authenticate once, and all 33+ tools are discovered automatically — email, vault, wallet, and compute in a single connection. No local software to install or maintain.

## Server URL

```
https://agent-loadout.com/api/mcp
```

## Quick connect

Add Agent Loadout to Claude Code in one command:

```bash title="Add to Claude Code" theme={null}
claude mcp add --transport http agent-loadout https://agent-loadout.com/api/mcp
```

After running this, Claude Code will prompt you to authenticate. Sign in to your Agent Loadout account in the browser, choose an agent, review its permissions, and approve. The connection is live in your next chat session.

## Supported clients

<CardGroup cols={2}>
  <Card title="Claude Code" icon="terminal">
    Add with `claude mcp add --transport http`. Works in the CLI and in the VS Code extension.
  </Card>

  <Card title="Codex CLI" icon="code">
    Configure the MCP server URL in your Codex profile. Authentication uses the same OAuth flow.
  </Card>

  <Card title="Cursor" icon="mouse-pointer">
    Add Agent Loadout as a custom MCP server in Cursor's settings. No plugin installation needed.
  </Card>

  <Card title="ChatGPT" icon="message">
    Enable Developer mode in ChatGPT settings, then add the server URL as a custom plugin connection using OAuth.
  </Card>
</CardGroup>

Any client that speaks MCP over HTTP with OAuth 2.1 will work. See the [OAuth page](/api-reference/mcp/oauth) for integration details.

## Same token, same capabilities

Each MCP tool maps directly to a REST endpoint. Use whichever interface is more convenient — the same agent token, the same scoped permissions, and the same audit trail apply to both. You never need to manage two separate sets of credentials.

## What happens after connecting

Once connected, your client discovers all available tools automatically. Which tools are active depends on the agent's granted permissions and your workspace's enabled features:

* **Email tools** require `email:read` and/or `email:send`
* **Vault tools** require `vault:metadata`, `vault:use`, and/or `vault:write`
* **Wallet tools** require `wallet:read` and/or `wallet:pay` (feature-flagged)
* **Compute tools** require `compute:read`, `compute:run`, and/or `compute:admin` (feature-flagged)

<Note>
  Access to a credential's value must be explicitly enabled on the credential itself. Wallet and machine features must also be available in your workspace plan.
</Note>

## Learn more

<CardGroup cols={2}>
  <Card title="MCP Tools Reference" icon="list" href="/api-reference/mcp/tools">
    Every tool, grouped by category, with scope requirements and descriptions.
  </Card>

  <Card title="MCP OAuth" icon="lock" href="/api-reference/mcp/oauth">
    Discovery documents, PKCE flow, token lifetimes, and scope definitions for client developers.
  </Card>
</CardGroup>