# Bagel

> Bagel is an open-source MCP server for analytics and data. Ask plain-English questions of robot, drone and IoT logs, with auditable SQL answers. It runs as a hosted remote server, so there is nothing to install locally.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** CLI, self-hosted, Linux, macOS
- **Website:** https://www.trybagel.com
- **Repository:** https://github.com/extelligence-ai/bagel
- **MCP Registry name:** io.github.Extelligence-ai/bagel
- **Categories:** analytics and data, developer tools
- **Listing updated:** 2026-11-16

## About

Bagel is built on one idea: language models are good at questions and unreliable at arithmetic, so the numbers come from DuckDB. The model picks the query, DuckDB runs it over an extract of your actual messages, and Bagel shows you the SQL so you can check it. A bundled PX4 sample, for instance, answers a lowest-battery-voltage question with the value and the time it occurred.

It reads ROS 1 and ROS 2 bags, PX4, ArduPilot and Betaflight flight logs, CAN and MDF4 files, CSV, JSON, Parquet, ROS text logs and live MQTT through optional extras. A second feature turns a sentence such as keeping ten seconds around every hard brake into a pipeline that is previewed before anything is written, then run once, across a fleet or standing on the robot, optionally uploading slices to S3, GCS or Azure.

The LLM stays in front of your logs and never in the robot's control loop. It installs from PyPI as bagel-mcp, or runs in Docker images for live robots and edge use, and works with Claude Code, Gemini, Cursor, Codex, Grok or a local model. The site flags its event-detection feature as beta.

## Features

- **SQL-backed answers:** Calculations run as DuckDB SQL that Bagel shows you, not as model guesses.
- **Wide format support:** Reads ROS bags, flight logs, CAN and MDF4, CSV, JSON, Parquet and live MQTT via extras.
- **Natural-language pipelines:** Describe an event and keep only the windows around it, previewed before writing.
- **Edge data reduction:** Pipelines can run on the robot and upload only the kept slices.
- **No-LLM demo:** A Docker demo prints a robot-health report card from a bundled sample log.
- **Model choice:** Works with several coding assistants or a fully offline model.

## Good for

- Find the lowest battery voltage and when it happened in a flight log
- Read ROS WARN and ERROR logs and explain a crash
- Check an IMU or odometry channel for unusual readings
- Cut hours of bag data down to the seconds around an incident
- Check recorded data for gaps

## Install

### Claude Code (Terminal)

```sh
claude mcp add bagel -- docker run -i --rm ghcr.io/extelligence-ai/bagel/ros2-kilted:2.4.1
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "bagel": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/extelligence-ai/bagel/ros2-kilted:2.4.1"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "bagel": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/extelligence-ai/bagel/ros2-kilted:2.4.1"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "bagel": {
      "type": "stdio",
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/extelligence-ai/bagel/ros2-kilted:2.4.1"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "bagel": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/extelligence-ai/bagel/ros2-kilted:2.4.1"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "bagel": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "ghcr.io/extelligence-ai/bagel/ros2-kilted:2.4.1"
      ]
    }
  }
}
```

## FAQ

### Is it open source?

Yes. The site lists Apache 2.0, and the free open-source release installs with pip install bagel-mcp.

### Do I need ROS installed?

Not for recorded data; the Docker images cover live robots and the edge.

### Are its numbers generated by the model?

No. Every calculation is DuckDB SQL, and the query is shown for audit.

### Which Python versions work?

Python 3.10 to 3.12 for the pip install.

### Is Bagel free?

Yes. Bagel is open source, so you can use and self-host it for free.

### How do I add Bagel to Claude Code?

Run this command in your terminal: claude mcp add bagel -- docker run -i --rm ghcr.io/extelligence-ai/bagel/ros2-kilted:2.4.1

### What are the best alternatives to Bagel?

The closest alternatives to Bagel are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

## Alternatives

- [REA](https://ai-powered.com/mcp/rea): REA is an open-source MCP server for developer tools.
- [Codebase Memory](https://ai-powered.com/mcp/codebase-memory): Codebase Memory is an open-source MCP server for knowledge and memory.
- [GitHub](https://ai-powered.com/mcp/github-2): GitHub is an open-source MCP server for developer tools.
- [Serena MCP: the IDE for your agent](https://ai-powered.com/mcp/serena-mcp-the-ide-for-your-agent): Serena MCP: the IDE for your agent is an open-source MCP server for developer tools.
- [XERJ](https://ai-powered.com/mcp/xerj): XERJ is an open-source MCP server for knowledge and memory.
- [Agent-Native Analytics](https://ai-powered.com/mcp/agent-native-analytics): Agent-Native Analytics is an open-source MCP server for analytics and data.
- [Ouroboros](https://ai-powered.com/mcp/ouroboros): Ouroboros is an open-source MCP server for developer tools.
- [Claude Code Ultimate Guide](https://ai-powered.com/mcp/claude-code-ultimate-guide): Claude Code Ultimate Guide is an open-source MCP server for ai models and llm tooling.

---
Source: https://ai-powered.com/mcp/bagel
