Sphero RVR MCP Server

An MCP (Model Context Protocol) server that enables AI assistants to control a Sphero RVR robot. Run this on a Raspberry Pi connected to your RVR, and use any MCP-compatible client to drive, control LEDs, read sensors, and more.

Features

Core Capabilities

• Full RVR Control: Movement, LEDs, sensors, battery monitoring, IR communication

• Distance-Based Movement: Drive forward/backward by meters, pivot by degrees

• Safety System: Configurable speed limits, auto-stop timeout, emergency stop

• Sensor Streaming: Background streaming with cached data access

• Natural Language Control: Let AI drive your robot with conversational commands

• Client Agnostic: Works with any MCP-compatible client

Low-Latency Architecture (v0.2.1+)

• Direct Serial Protocol: Bypasses SDK for sub-millisecond command latency

• No Sphero SDK Required: Works on Python 3.10 - 3.13+ (SDK was limited to 3.10)

• Minimal Dependencies: Just 4 packages (fastmcp, pydantic, structlog, pyserial)

• Distance Control: drive_forward(0.5) moves exactly 0.5 meters

• Angle Control: pivot(90) rotates exactly 90 degrees

• Command Queue: Priority-based async queue eliminates race conditions

• Atomic State Management: Thread-safe state with validated transitions

• Structured Logging: JSON or console format for easy debugging

Related MCP server: VectorClaw

Compatible MCP Clients

• Claude Code (CLI)

• Claude Desktop

• Cursor

• Zed

• Continue.dev

• Any custom client using the MCP SDK

Requirements

• Raspberry Pi 3 or newer (connected to Sphero RVR via serial)

• Python 3.10+ (tested on 3.10, 3.12, 3.13 - no SDK limitations)

• Sphero RVR with serial connection to Pi

Installation

Install from PyPI (recommended)

pip install sphero-rvr-mcp

Install from source

git clone https://github.com/jsperson/sphero_rvr_mcp.git
cd sphero_rvr_mcp
pip install -e .

Verify Installation

Run the pre-flight check to verify everything is set up correctly:

sphero-rvr-mcp --check

This will verify:

• Python version (requires 3.10+)

• FastMCP is installed

• Serial port exists and is accessible

• Current configuration settings

Configure Your MCP Client

The server runs via stdio. Configure your MCP client with:

• Command: python -m sphero_rvr_mcp

Claude Code

claude mcp add sphero-rvr -c "python -m sphero_rvr_mcp"

Or edit ~/.claude.json:

{
  "mcpServers": {
    "sphero-rvr": {
      "type": "stdio",
      "command": "python",
      "args": ["-m", "sphero_rvr_mcp"]
    }
  }
}

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "sphero-rvr": {
      "command": "python",
      "args": ["-m", "sphero_rvr_mcp"]
    }
  }
}

Other Clients

Refer to your client's MCP configuration documentation. The server uses stdio transport.

Configuration

Environment variables (optional):

Usage

Example Commands

Once your MCP client is connected to the server:

You: Connect to the RVR

You: Drive forward 6 inches

You: Pivot 90 degrees to the right

You: Drive forward 1 meter

You: Set all LEDs to blue

You: What's the battery level?

You: Which direction is the rover facing?
# Returns heading in degrees and cardinal direction (N, NE, E, SE, S, SW, W, NW)

You: Turn left 45 degrees and drive backward 0.5 meters

You: Emergency stop!

Available Tools

Connection (3 tools)

Movement (11 tools)

LEDs (3 tools)

LED groups: headlight_left, headlight_right, brakelight_left, brakelight_right, status_indication_left, status_indication_right, battery_door_front, battery_door_rear, power_button_front, power_button_rear, undercarriage_white, all

Sensors (12 tools)

Streamable sensors: accelerometer, gyroscope, imu, locator, velocity, speed, quaternion, color_detection, ambient_light, core_time

Battery & System (11 tools)

Safety & Motor Protection (5 tools)

IR Communication (7 tools)

Architecture

Component Overview

MCP Tool Handlers (FastMCP)
    |
    v
Direct Serial Protocol (low-latency)
    |
    +-- Packet Builder (commands.py)
    +-- Serial Connection (direct_serial.py)
    |
    v
Sphero RVR (via /dev/ttyAMA0)

Key Design Patterns

Direct Serial Protocol

• Bypasses the Sphero SDK for minimal latency

• Constructs raw SOP/EOP packets directly

• Sub-millisecond command transmission

• Supports all core RVR commands

Distance-Based Movement

• drive_forward(distance) - Uses RVR's internal position controller

• drive_backward(distance) - Accurate reverse movement

• pivot(degrees) - Precise rotation using heading control

Command Queue

• All hardware commands go through async priority queue

• Priority levels: EMERGENCY (0) -> HIGH (1) -> NORMAL (2) -> LOW (3)

• Per-command timeout enforcement

• Eliminates race conditions in concurrent access

Atomic State Management

• Thread-safe with explicit locks

• Validated state transitions

• All changes logged

Observability

Structured Logging

{
  "event": "command_submitted",
  "command_type": "drive_forward",
  "distance": 0.5,
  "timestamp": "2026-01-15T01:23:45.678Z"
}

Safety Features

Speed Limiting

All movement commands are limited to a configurable percentage of max speed (default 50%). This prevents accidental high-speed collisions.

You: Set the speed limit to 25%
You: Now drive forward at full speed
# RVR will only go at 25% of max speed

Command Timeout

If no movement command is received within the timeout period (default 5 seconds), the RVR automatically stops. This prevents runaway situations if connection is lost.

Emergency Stop

Immediately stops all movement and blocks further motion until explicitly cleared. Has highest priority in the command queue.

You: Emergency stop!
# RVR stops immediately
# All movement commands will fail until:
You: Clear the emergency stop

Troubleshooting

Connection Issues

"Failed to connect to RVR"

• Check serial connection: ls -l /dev/ttyAMA0

• Ensure RVR is powered on and charged

• Verify baud rate (default 115200)

Connection times out

• RVR might be off or in deep sleep

• Try power cycling the RVR

• Check serial cable connection

Performance Issues

Slow response

• Raspberry Pi 3 may be slower than Pi 4/5

• Reduce sensor streaming frequency

• Close unnecessary applications

Sensor Issues

Color detection returns all zeros

• Ensure RVR is on a non-dark surface

• Try increasing stabilization time

• Check that belly LED is working

RVR Behavior Issues

RVR not responding to commands

• Check if emergency stop is active: get_safety_status

• Check speed limit isn't set to 0%

• Try disconnecting and reconnecting

RVR stops on its own

• Check command timeout setting (default 5 seconds)

• Disable auto-stop: set_command_timeout(0)

• Verify battery level isn't critical

Project Structure

sphero_rvr_mcp/
├── pyproject.toml              # Package configuration
├── README.md                   # This file
├── LICENSE                     # MIT License
└── src/sphero_rvr_mcp/
    ├── __init__.py
    ├── __main__.py             # Entry point
    ├── config.py               # Configuration
    ├── api.py                  # Direct API (non-MCP)
    ├── server.py               # MCP server & tool handlers
    │
    ├── protocol/               # Direct serial protocol
    │   ├── __init__.py
    │   ├── commands.py         # Packet builders
    │   ├── direct_serial.py    # Serial connection
    │   └── packet.py           # SOP/EOP framing
    │
    ├── core/                   # Core infrastructure
    │   ├── command_queue.py    # Priority command queue
    │   ├── state_manager.py    # Atomic state management
    │   └── exceptions.py       # Exception hierarchy
    │
    ├── hardware/               # Hardware abstraction
    │   ├── connection_manager.py       # Connection lifecycle
    │   ├── sensor_stream_manager.py    # Sensor streaming
    │   └── safety_monitor.py           # Safety system
    │
    ├── services/               # Application services
    │   ├── movement_service.py
    │   ├── led_service.py
    │   └── ir_service.py
    │
    └── observability/          # Logging
        └── logging.py          # Structured logging

Development

Running Tests

# Unit tests
pytest tests/unit/

# Integration tests (requires RVR)
pytest tests/integration/

# All tests
pytest

Viewing Logs

# Console format (human-readable)
RVR_LOG_FORMAT=console python -m sphero_rvr_mcp

# JSON format (machine-readable)
RVR_LOG_FORMAT=json python -m sphero_rvr_mcp

Contributing

Contributions welcome! Please:

1. Fork the repository

2. Create a feature branch

3. Add tests for new functionality

4. Ensure all tests pass

5. Submit a pull request

Changelog

v0.2.3 (2026-01-29)

• 22 new MCP tools for full SDK sensor coverage:

  • Temperature & thermal protection monitoring

  • System info (firmware, MAC, SKU, uptime, board revision)

  • Extended battery info (voltage, state, thresholds)

  • Motion sensors (encoders, magnetometer with compass heading)

  • Motor protection (fault state, stall/fault notifications)

  • IR follow/evade behaviors

• Fixed 4 sensor commands with incorrect protocol parameters

• Magnetometer heading - returns heading (degrees) and cardinal direction

v0.2.1 (2026-01-15)

• Direct serial protocol for low-latency control (bypasses SDK)

• Added drive_forward and drive_backward with distance in meters

• Added pivot command for precise angle rotation

• Removed unused dependencies (prometheus-client, tenacity, nest-asyncio)

• Added pyserial dependency

• Simplified architecture (removed circuit breaker, event bus, metrics)

• No longer requires Sphero SDK installation

v0.2.0 (2026-01-14)

• Complete architectural rewrite for production reliability

• Added command queue with priority levels

• Added circuit breaker for connection resilience

• Added event bus for sensor distribution

• Added atomic state management

• Added comprehensive observability (logging + metrics)

• Fixed SDK response key parsing (color sensor now works!)

v0.1.1 (2024-12-XX)

• Added connection timeouts

• Updated documentation

v0.1.0 (2024-12-XX)

• Initial release

License

MIT

Credits

• FastMCP

• Model Context Protocol

• Sphero RVR