Fleetctl MCP integration

Name: Fleetctl
Brand: Fleetctl
Author: Fleetctl

Fleetctl exposes 26 local tools via the Model Context Protocol (MCP), plus hosted Streamable HTTP MCP at https://app.fleetctl.ai/mcp. The server card is /.well-known/mcp/server-card.json. Clients discover local tools from the running binary, so the binary remains the source of truth.

Browse the marketing solution pages for all 26 tools

This page documents Fleet's local MCP surface for agents, tasks, events, settings, and diagnostics. Any compatible MCP client can also connect to Fleet's hosted control plane at https://app.fleetctl.ai/mcpfor workflow definitions, approvals, and run history, using Fleet's OAuth sign-in.See conversational workflow authoring. The removed fleet_pipeline_* tools are not part of the local surface.

Setup

Add to Cursor

One click writes https://app.fleetctl.ai/mcp into ~/.cursor/mcp.json. Cursor then prompts for Fleet OAuth. Same result from the CLI: fleet mcp install.

Add Fleet MCP to Cursor

1. Choose a transport

Hosted Streamable HTTP (OAuth, no local binary):

https://app.fleetctl.ai/mcp

Local stdio using JSON-RPC 2.0 (after the binary is installed):

fleet mcp serve

2. Configure .mcp.json

Add this to your project root so Claude Code auto-discovers the server. Codex, OpenCode, and other MCP clients use the same fleet mcp serve command or the hosted Streamable HTTP URL:

{
  "mcpServers": {
    "fleet": {
      "command": "fleet",
      "args": ["mcp", "serve"],
      "env": {}
    }
  }
}

How it works

When an MCP client starts, it launches fleet mcp serve as a subprocess and communicates via JSON-RPC 2.0 over stdin/stdout. All 26 tools appear as available tools in that session. No network ports, no HTTP — just stdio. Hosted Streamable HTTP is the same protocol without a local subprocess.

Fleet & Status Tools

5 tools. Inspect health, diagnostics, logs, dashboards, and Brain insights.

Tool	Description	Solution
fleet_status	Get agent counts, pending deliveries, and Brain alerts.	View page
fleet_doctor	Check tmux, SQLite, data paths, sockets, and GitHub CLI.	View page
fleet_log	Read the unified Fabric and agent activity timeline.	View page
fleet_dashboard	Generate a CEO, CTO, CPO, CMO, or general operating view.	View page
fleet_brain_insights	Return evaluation, risk, thoughts, and recommendations.	View page

Agent Tools

7 tools. Create, inspect, start, stop, and remove Fleet-managed agents.

Tool	Description	Solution
fleet_agent_list	List agents with status, vendor, model, and mode.	View page
fleet_agent_start	Start an enabled, stopped agent by fuzzy name.	View page
fleet_agent_stop	Stop a running agent by fuzzy name.	View page
fleet_agent_create	Create a configured agent instance.	View page
fleet_agent_delete	Delete an agent, optionally stopping it first.	View page
fleet_agent_logs	Return raw agent log lines.	View page
fleet_agent_output	Return run times, exit codes, durations, PRs, and commits.	View page

Fabric Tools

2 tools. Publish coordination events and inspect pending inbox deliveries.

Tool	Description	Solution
fleet_fabric_publish	Publish a task, handoff, blocker, decision, update, request, or review event.	View page
fleet_fabric_inbox	List deliveries by agent and status.	View page

Watch Tools

2 tools. Notify an agent when a watched GitHub issue closes.

Tool	Description	Solution
fleet_watch_create	Register a project-scoped issue-close watch.	View page
fleet_watch_list	List active watches, optionally by agent.	View page

Task & Context Tools

5 tools. Assign work and inspect or update Fabric task state.

Tool	Description	Solution
fleet_agent_context	Return the agent config and pending inbox.	View page
fleet_task_assign	Set FLEET_TASK_TITLE and start the selected agent immediately.	View page
fleet_task_get	Return title, description, status, priority, and assignee.	View page
fleet_task_list	List tasks, optionally filtered by status.	View page
fleet_task_update	Change task status or assignee.	View page

Configuration Tools

3 tools. Read and update local Fleet settings.

Tool	Description	Solution
fleet_config_list	List configuration as key/value pairs.	View page
fleet_config_get	Read one setting by key.	View page
fleet_config_set	Write one setting key and value.	View page

Workflow Tools

2 tools. Inspect and update saved control-plane workflow execution settings.

Tool	Description	Solution
fleet_workflow_get	Read a saved control-plane workflow by its opaque ID.	View page
fleet_workflow_update	Update a saved workflow's runner and model without replacing its definition.	View page

Conversational Examples

With MCP configured, you can manage your fleet through natural conversation in Claude Code, Codex, OpenCode, or any other MCP client:

"Show me the fleet status"

Uses: fleet_status

Returns color-coded health summary of all agents, watchers, and pipelines.

"Start the frontend-dev agent and assign it the dark mode task"

Uses: fleet_agent_start + fleet_task_assign

Starts the agent in a tmux session, then dispatches the task with FLEET_TASK_TITLE.

"What has the QA team been doing in the last hour?"

Uses: fleet_log

Queries the unified fleet log filtered by team and time range.

"Assign issue #42 to backend-dev"

Uses: fleet_task_assign

Starts the selected agent with the task as authoritative context.

"Check if frontend-dev has any inbox messages"

Uses: fleet_fabric_inbox

Returns unread fabric inbox messages for the agent.

Protocol Details

Transport

JSON-RPC 2.0 over stdio (stdin/stdout). No HTTP server, no WebSocket, no network ports. The MCP client spawns fleet mcp serve as a child process and communicates via pipes.

Request Format

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "fleet_agent_start",
    "arguments": {
      "name": "frontend-dev"
    }
  }
}

Response Format

{
  "jsonrpc": "2.0",
  "id": 1,
  "result": {
    "content": [
      {
        "type": "text",
        "text": "Agent frontend-dev started successfully"
      }
    ]
  }
}

Discovery

The MCP server responds to tools/list with all 26 tool definitions including parameter schemas. Clients call this on startup to populate their tool list.