MCP Server — Connect AI Agents Directly

FleetQ exposes a Model Context Protocol (MCP) server that gives LLMs and agent frameworks direct, tool-based access to every platform capability. No REST calls, no token management in your prompts — just natural language.

Scenario: A developer connects Claude Code to FleetQ via MCP and types: "Create an experiment that monitors Hacker News for mentions of our product and summarises the top 5 posts every morning." Claude calls the agent_create, skill_create, and project_create MCP tools in sequence — the entire setup is done in one conversation.

Three endpoints, two servers

FleetQ ships a compact and a full MCP server so clients with tool-count limits (Claude.ai) and clients without (Cursor, Claude Code) can both use the platform natively:

Both HTTP servers share the same OAuth discovery endpoints, consent screen, and BYOK provider credentials. The compact server consolidates related operations into 34 meta-tools that take an action parameter (e.g. agent_manage with action=list), delegating internally to the same 400 implementations used by the full server.

Setting up stdio (Claude Code)

Add FleetQ to your Claude Code MCP configuration:

{
  "mcp_servers": {
    "agent-fleet": {
      "command": "php",
      "args": ["/path/to/agent-fleet/artisan", "mcp:start", "agent-fleet"]
    }
  }
}

Or start it directly from the terminal:

php artisan mcp:start agent-fleet

Setting up HTTP (Cursor, remote)

Create an API token in Team Settings → API Tokens, then configure your MCP client:

{
  "mcpServers": {
    "agent-fleet": {
      "url": "https://fleetq.169.58.89.204.sslip.io/mcp/full",
      "headers": {
        "Authorization": "Bearer YOUR_TOKEN"
      }
    }
  }
}

For Claude.ai and other clients with tool-count limits, use https://fleetq.169.58.89.204.sslip.io/mcp (compact, 34 meta-tools) instead of /mcp/full.

Setting up in Claude.ai (OAuth)

Claude.ai, Claude Desktop, and other MCP clients that speak OAuth 2.0 can connect without a pre-issued API token. Go to Settings → Connectors → Add custom connector in Claude.ai and paste the server URL:

https://fleetq.169.58.89.204.sslip.io/mcp

FleetQ implements the full MCP authorization spec:

• Authorization Code flow with PKCE (S256) — no client secret required for public clients
• Dynamic Client Registration (RFC 7591) at /oauth/register — no manual client setup
• Authorization Server metadata (RFC 8414) at /.well-known/oauth-authorization-server
• Protected Resource metadata (RFC 9728) at /.well-known/oauth-protected-resource
• Scope: mcp:use (single scope covering all tool invocations)
• Refresh tokens via the standard refresh_token grant

When Claude.ai connects, you will be redirected to the FleetQ consent screen to authorize access for the calling team. Tokens are scoped to the team you choose at consent time — no cross-tenant access is possible.

Tool profiles (compact server only)

On the compact server (/mcp), teams can further restrict which meta-tools are visible via profiles in Team Settings → MCP Tools. The full server (/mcp/full) always exposes all 400+ tools.

34 meta-tools on /mcp (compact server)

This is what Claude.ai, Claude Desktop, and other tool-limited clients see. Each meta-tool takes an action parameter plus action-specific arguments, and delegates internally to the corresponding granular tool on the full server:

Need direct access to each granular operation? Connect to /mcp/full instead — it exposes 400+ individual tools across 33 domains (e.g. agent_list, agent_create, agent_update as separate entries in tools/list). Both servers share the same OAuth flow, BYOK credentials, and ApprovalsResource MCP App widget.

Troubleshooting

Common issues and fixes when connecting the FleetQ MCP server:

Still stuck? Email support@fleetq.net with your team ID (visible in Team Settings) and a description of what you were trying to do. We respond within 2 business hours during review.

Privacy, terms & support

Use of the FleetQ MCP server is governed by our public policies:

• Privacy Policy — GDPR + CCPA compliant, covers data collection, retention, third-party sharing, and your rights.
• Terms of Service — acceptable use, liability, and service level.
• Support: support@fleetq.net for connector issues, onboarding questions, or enterprise enquiries.