{"id":"f5ac430f-4a64-490b-9f6f-3a191fdb7c8d","slug":"duckdb-eda-mcp-server","kind":"mcp","name":"duckdb-eda-mcp-server","summary":"MCP server for Exploratory Data Analytics using DuckDB","owner":"thedevtoolsmith","repoUrl":"https://github.com/thedevtoolsmith/duckdb-eda-mcp-server","license":"GPL-3.0","licenseTier":"strong-copyleft","trust":"community","stars":1,"match":null,"fewStars":true,"format":"mcp","createdAt":"2025-04-07T16:46:50.000Z","addedAt":"2026-10-21T14:58:22.401Z","logoUrl":"https://avatars.githubusercontent.com/u/29922252?u=49ab694effbf6c49396d1c17e66483b6519f7938&v=4","linkOnly":false,"collection":null,"skillPath":null,"categories":["databases","cloud-infra","data-analytics"],"tags":[],"runtime":"python","local":true,"cloudCapable":false,"cloudCapability":"unknown","cloudReadyOnPlatform":false,"needsApiKey":false,"reviewed":false,"syncedAt":"2026-10-24T00:02:05.114Z","sample":false,"url":"https://conduits.dev/tools/duckdb-eda-mcp-server","language":"Python","transports":["stdio"],"install":{"pypi":"dems"},"requirements":{"env":[],"upstreamMayCharge":false},"tools":null,"readme":"# DuckDB EDA MCP Server (DEMS)\n\n## About\n\nDEMS is intended to be helpful for anyone who wants to do exploratory data analytics, i.e., do some quick and dirty\nanalytics of some data to present an argument or to prove a point using DuckDB. (If you're not using DuckDB, [check it\nout](https://duckdb.org/). It's a *really* fast columnar, in-process DB)\n\nYou might be a developer/product manager who wants to make a business case for building a new feature, or somebody\nimporting and analysing some data for writing a blog post. This MCP server is designed to help you quickly sift through\nthe data.\n\n## Features\n\n- Tool to help import data from CSV and JSON files. *Don't spend valuable credits asking LLMs to generate queries to\n  import data.*\n- DELETE, DROP and UPDATE queries are automatically blocked; *LLMs cannot mess up your data “accidentally.”*\n- Tool to validate SQL queries; *This helps LLMs know when it generates garbage.*\n- Tools that are tailored for exploratory data analytics using DuckDB like `get_sample_data`,\n  `get_table_schema_and_stats`. *Helps the LLM leverage DuckDB capabilities instead of generating the same SQL over and\n  over again.*\n- Default timeout of 60 seconds for all queried. *This will save you if LLMs manage to mess up in spite of all the other\n  safeguards.*\n\n## Server Installation\n\nThis MCP has to be installed and run locally, as DuckDB will be present in your file system (Data privacy is an\nunintended side effect). The recommended way to run this locally is using `uv`\n\n```shell\nuvx --from git+https://github.com/thedevtoolsmith/duckdb-eda-mcp-server dems --db <DUCKDB_FILE_PATH>\n```\n\n## Integration with Client\n\nThe MCP server can be integrated into an MCP client by copying and pasting the following config block into your config\nfile:\n\n```json\n{\n  \"mcpServers\": {\n    \"dems\": {\n      \"command\": \"npx\",\n      \"args\": [\n        \"mcp-remote\",\n        \"http://localhost:8000/sse\"\n      ]\n    }\n  }\n}\n```\n\n## Roadmap\n\nIf there is enough interest, the following features are planned:\n\n- [ ] Add the DuckDB documentation as a resource\n- [ ] Add tools to give a column summary\n- [ ] Add prompts to help users with EDA-related use cases"}