SQLite MCP Server
## what it does
tl;dr: archived reference server. Source is frozen at modelcontextprotocol/servers-archived/src/sqlite. npx still resolves the npm package, but the archive README is explicit: NO SECURITY GUARANTEES ARE PROVIDED FOR THESE ARCHIVED SERVERS.
Status: archived
SQLite was moved out of the active monorepo in May 2025 when the MCP steering group reorganized maintenance. The archive repo — modelcontextprotocol/servers-archived — is itself marked archived on GitHub and no longer accepts contributions.
What that means in practice:
- Source is frozen at
servers-archived/src/sqlite. - The npm package
@modelcontextprotocol/server-sqlitestill resolves fromnpx, but receives no security updates and no bug fixes. - The archive README lists thirteen servers as historical reference only: AWS KB Retrieval, Brave Search, EverArt, Git, GitHub, GitLab, Google Drive, Google Maps, PostgreSQL, Puppeteer, Redis, Sentry, Slack, SQLite.
For a maintained SQLite bridge, use the project’s canonical aggregator: the MCP Registry. Some projects (e.g. Slack → zencoderai/slack-mcp-server) have been forked and are now maintained externally; SQLite does not have an official successor listed.
What it does
The SQLite MCP server connects your agent to a local SQLite database file. The agent can inspect the schema, run SELECT queries for analysis, and execute INSERT/UPDATE/DELETE for data modification.
A unique feature is the append_insight tool — the server maintains a memo of analytical insights the agent discovers, building a running summary of what it has learned about the dataset.
Installation (legacy)
# Point it at your .db file:
npx -y @modelcontextprotocol/server-sqlite ~/data/myapp.db
# Or install globally:
npm install -g @modelcontextprotocol/server-sqlite
These commands resolve against the archived npm release. Treat the install as frozen at its published version.
Claude Desktop configuration
{
"mcpServers": {
"sqlite": {
"command": "npx",
"args": ["-y", "@modelcontextprotocol/server-sqlite", "/Users/you/data/myapp.db"]
}
}
}
Available tools
| Tool | Description |
|---|---|
read_query | Execute a SELECT — returns rows as JSON |
write_query | Execute INSERT/UPDATE/DELETE/CREATE |
create_table | Create a new table with schema |
list_tables | List all tables in the database |
describe_table | Get column names and types for a table |
append_insight | Log an analytical finding to the memo |
Practical use cases
- Data exploration: “What are the top 10 users by order count?”
- Schema inspection: “What tables exist and how are they related?”
- Reporting: “Give me monthly revenue by product category”
- Migration assistance: “Here’s the schema — help me write the migration”
Gotchas
- No security guarantees — the archive README spells this out in all caps.
- One database file per server instance. Multiple databases means multiple server processes.
- Write ops don’t auto-rollback. The agent should wrap multi-step modifications in a transaction.
- Cursor compatibility is partial — only the read tools work in some Cursor versions.
- The monorepo’s old
src/sqlitepath inmodelcontextprotocol/servers404s now; the canonical source isservers-archived.
Related
- MCP Registry — find a maintained SQLite bridge plus other data sources.
- PostgreSQL MCP — same fate: archived in
servers-archived/src/postgres. - GitHub MCP — the maintained successor for the GitHub reference server that was also archived.
Stars (296) and forks (166) reflect the modelcontextprotocol/servers-archived repo as fetched this session. installs is carried over from the prior entry. Sandbox test not run.
## compatibility
## exposed tools
## embed this badge
