MCP ROS Bags

A Model Context Protocol (MCP) server for analyzing and querying ROS bag files. This server enables Claude and other MCP-compatible tools to interact with robot data stored in ROS bags.

Overview

This MCP server provides comprehensive tools for ROS bag analysis, supporting both ROS 1 and ROS 2 bag formats (.bag, .db3, .mcap). It enables AI assistants to query robot trajectories, sensor data, transforms, and logging information from recorded robot sessions.

Key Features

• Bag Management: List, inspect, and query ROS bag files
• Message Querying: Get messages by topic, timestamp, or time range
• Trajectory Analysis: Extract robot paths, waypoints, and motion profiles
• Sensor Data: Analyze laserscan topics, images
• Transform Trees: Query coordinate transforms and frame relationships
• Logging Analysis: Search and analyze ROS log messages
• Visualization: Generate plots and visual representations of data
• Search & Discovery: Find objects, patterns, and correlations in data

Installation

1. Clone the repository:

   git clone https://github.com/binabik-ai/mcp-rosbags.git
   cd mcp-rosbags
   

2. Create and activate a Python virtual environment:

   python3 -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   

3. Install dependencies:

   pip install -r requirements.txt
   

4. Optional: Install with ROS 2 support:

   pip install -e .[ros2]
   

Configuration

Claude Desktop Configuration

Add this to your Claude Desktop configuration file:

{
  "mcpServers": {
    "rosbag_reader": {
      "command": "/path/to/your/venv/bin/python",
      "args": ["/path/to/mcp-rosbags/src/server.py"],
      "env": {
        "PYTHONPATH": "/path/to/mcp-rosbags/src",
        "MCP_ROSBAG_DIR": "/path/to/your/rosbags",
        "MCP_ROSBAG_CONFIG": "/path/to/mcp-rosbags/src/config"
      }
    }
  }
}

Environment Variables

• MCP_ROSBAG_DIR: Default directory containing ROS bag files (default: ./rosbags)
• MCP_ROSBAG_CONFIG: Configuration directory (default: ./config)

Server Configuration

The server uses configuration files in the src/config/ directory:

• server_config.yaml: Main server configuration (caching, timeouts, etc.)
• message_schemas.yaml: Schema definitions for message field extraction

Usage

With Claude Desktop

After configuration, restart Claude Desktop and you can interact with your ROS bags:

"List all ROS bags in my directory"
"Show me the trajectory from the /odom topic"
"Find all error messages in the logs"
"Extract an image from the /camera/image topic at timestamp 1234567890 and describe what you see"

Tools Overview

The server provides the following categories of tools:

Core Tools

• set_bag_path: Set current bag file or directory
• list_bags: List available ROS bag files
• bag_info: Get detailed bag information
• get_message_at_time: Get message at specific timestamp
• get_messages_in_range: Get messages within time range

Analysis Tools

• Trajectory: Extract robot paths, analyze motion patterns
• LiDAR: Process laserscan data, e.g., to detect obstacles
• Images: Extract and process camera images
• Transforms: Query coordinate transforms and frame trees
• Logging: Search and analyze ROS log messages
• Visualization: Generate plots and visual data representations

Search Tools

• Pattern matching in message data
• Correlation analysis between topics
• Time-based data alignment

Architecture

src/
├── server.py              # Main MCP server
├── config/                # Configuration files
│   ├── server_config.yaml # Server settings
│   └── message_schemas.yaml # Message field schemas
├── core/                  # Core utilities
│   ├── cache_manager.py   # Caching system
│   ├── message_utils.py   # Message processing
│   └── schema_manager.py  # Schema handling
└── extractors/           # Specialized analysis tools
    ├── bag_management.py # Bag operations
    ├── image.py          # Image processing
    ├── lidar.py          # LiDAR analysis
    ├── logging.py        # Log analysis
    ├── search.py         # Search and correlation
    ├── tf_tree.py        # Transform trees
    ├── trajectory.py     # Path analysis
    └── visualization.py  # Data visualization

Supported Message Types

The server supports standard ROS message types including:

• Navigation: nav_msgs/Odometry, nav_msgs/Path
• Geometry: geometry_msgs/PoseStamped, geometry_msgs/Twist
• Sensors: sensor_msgs/LaserScan, sensor_msgs/PointCloud2, sensor_msgs/Image
• Transforms: tf2_msgs/TFMessage, geometry_msgs/TransformStamped
• Diagnostics: diagnostic_msgs/DiagnosticArray
• Logging: rcl_interfaces/Log

Performance Features

• Intelligent Caching: Configurable caching system for frequently accessed data
• Schema-based Extraction: Efficient field extraction using predefined schemas

License

This project is licensed under the Apache License 2.0. See LICENSE for details.

Support

• Issues: Report bugs and request features on GitHub Issues
• Contact: dev@binabik.ai

Acknowledgments

Built with the Model Context Protocol and the rosbags library.