home / mcp / spotdb mcp server

SpotDB MCP Server

Provides a lightweight data sandbox with MCP and REST interfaces for safe AI-driven data exploration.

Installation
Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "aliengiraffe-spotdb": {
      "url": "http://localhost:8081/stream"
    }
  }
}

SpotDB provides a lightweight, ephemeral data sandbox that lets you explore data with AI while enforcing guardrails and security. It exposes data through an MCP interface and a REST API, enabling safe, isolated analysis without touching production databases.

How to use

You can use SpotDB by running the local server, uploading data, and querying it through either the MCP interface or the REST API. You can also connect an MCP client (like Claude Code) to run data analysis workflows against your sandbox. Use the Explorer UI to upload files and run queries in a browser.

Typical usage patterns include starting SpotDB, adding a CSV as a table, and performing SQL-like queries to inspect results. You can also set up an MCP integration to drive AI-powered analysis and keep data isolated from production systems.

How to install

Prerequisites: you need a Unix-like environment with a shell, Homebrew for macOS users, and network access to run local commands.

# Install the SpotDB package via Homebrew
brew tap aliengiraffe/spaceship && \\ 
brew install spotdb

# Start the SpotDB server
spotdb

# Optional: upload a CSV file to create a table
curl -X POST \
  http://localhost:8080/api/v1/upload \
  -F "table_name=mytable" \
  -F "has_header=true" \
  -F "[email protected]"

# Optional: query the data
curl -X POST \
  http://localhost:8080/api/v1/query \
  -H "Content-Type: application/json" \
  -d '{"query": "SELECT * FROM mytable LIMIT 10"}'

# Optional: connect an MCP client (example with Claude Code)
claude mcp add spotdb -s user -- npx -y mcp-remote http://localhost:8081/stream

# Open the Explorer UI
open http://localhost:8080/explorer

Configuration and usage notes

SpotDB can be interacted with in two main ways: via the MCP channel for AI models and workflows, or via the REST API for traditional integrations. The REST API exposes endpoints to upload data and run queries, while the MCP path lets you hook SpotDB into AI agents and automated pipelines.

You will typically run SpotDB locally and connect your MCP client or UI to the provided local endpoints. The REST API endpoint for uploads is http://localhost:8080/api/v1/upload and the query endpoint is http://localhost:8080/api/v1/query. The Explorer UI is accessible at http://localhost:8080/explorer.

MCP server configuration

The following MCP configurations enable remote and local access to SpotDB for MCP-based workflows. The HTTP-based MCP URL is used for remote clients, and a local stdio command starts SpotDB directly.

{
  "mcpServers": [
    {
      "type": "http",
      "name": "spotdb_http",
      "url": "http://localhost:8081/stream",
      "args": []
    },
    {
      "type": "stdio",
      "name": "spotdb",
      "command": "spotdb",
      "args": []
    }
  ]
}

Security and guardrails

SpotDB is designed to prevent accidental changes to production data. It enforces guardrails to limit data exposure and maintain privacy during AI-driven analysis. Run in an isolated sandbox and avoid direct access to production databases when using AI agents.

Troubleshooting

If you cannot access the REST API, verify that SpotDB is running and reachable at http://localhost:8080. If using MCP, ensure the remote MCP URL is reachable (for example, http://localhost:8081/stream) and that your MCP client is configured to connect to SpotDB.

Notes

This sandbox supports ephemeral databases, capture of data snapshots, REST API access, and MCP API access for AI workflows. Use the Explorer UI to interact with data visually and keep operations isolated from production systems.

Available tools

upload_csv

Upload a CSV file to create a new table in the sandbox using the REST API.

query_sql

Execute SQL-like queries against the uploaded data via the REST API.

snapshot

Capture and manage point-in-time data snapshots for recovery and state continuation.

mcp_integration

Integrate SpotDB with MCP workflows by linking an MCP client to the local sandbox.

explorer_ui

Open and use the web-based Explorer UI to upload files and run queries.