MCP SQLite (jparker)

This is the MCP server you want if you're working with SQLite databases in Claude. It wraps the entire SQLite API into a clean set of tools: create tables, insert rows, run arbitrary queries, inspect schemas, even export to CSV. No ceremony, no middleware layer to fight with. You point it at a .db file and it just works. I've used this for two main workflows. First, prototyping data models before committing to Postgres. I can sketch out tables, seed them with realistic data, and have Claude run analytical queries to sanity-check my schema design. Second, inspecting unfamiliar SQLite databases. Instead of opening DB Browser or writing throwaway Python scripts, I ask Claude to show me the schema, sample some rows, and explain what's going on. It's faster than context-switching to another tool. The install is trivial: npx points at your database file, add the config block to Claude Desktop, restart. No environment variables, no auth dance. It's read-write by default, which is what you want for prototyping but means you should point it at a copy if you're inspecting production data. One quirk: it doesn't handle concurrent writes gracefully because SQLite doesn't. If you're running this alongside another process that's writing to the same file, you'll hit lock errors. Not the server's fault, just SQLite being SQLite. Also, there's no built-in backup mechanism, so if Claude decides to DROP TABLE based on a misunderstood prompt, you're on your own. Keep backups. The query tool is particularly good. You can ask Claude to write complex JOINs, aggregations, or window functions and it'll execute them directly. I've used this to generate summary reports from messy CSV imports without writing a single line of SQL myself. The schema inspection tool is equally useful: it returns CREATE TABLE statements, which Claude can then modify and re-run to evolve your schema. This isn't for production workloads or multi-user scenarios. It's for local development, data exploration, and throwaway prototypes. If you're building a real application, you'll outgrow SQLite quickly. But for everything up to that point, this server is the fastest path from idea to working database.