MCP Server

scripts/drone_mcp_server.py wraps drone_control.py as an MCP (Model Context Protocol) tool server. AI agents see native tool calls instead of CLI commands.

Available Tools

arm()                          → arm the drone
disarm()                       → disarm the drone
get_state()                    → full telemetry
set_offboard()                 → enter offboard mode
set_position(x, y, z, yaw)    → fly to position
takeoff()                      → autonomous takeoff
land()                         → land at current position
return_to_launch()             → RTL
set_target_drone("drone2")     → switch target drone
get_fleet_status()             → scan all drones
upload_mission(waypoints)      → upload waypoint mission
mission_control(command)       → START, PAUSE, RESUME, STOP
get_mission_status()           → mission progress
flight_termination()           → EMERGENCY motor cutoff

Setup

npm i -g mcporter
mcporter config add drone-control \
  --command '/path/to/uv run /path/to/drone-os/scripts/drone_mcp_server.py'

# Verify tools
mcporter list drone-control --schema

# Test a call
mcporter call drone-control.get_state()

Configuration

Config lives in .mcp.json in the repo root. For MCP-compatible agent gateways, register via mcporter.

Relationship to Other Interfaces

drone_mcp_server.py imports drone_control.py as a library — it's a thin wrapper that exposes the same functions as MCP tools. No duplicate logic.

The MCP server is not used by the current dispatch setup (which uses the droneos CLI directly). It exists for compatibility with AI agents and tools that support MCP.

Available Tools

Setup

Configuration

Relationship to Other Interfaces

On this page