sqlite-mcp

Expert MCP for maximal, correct, and efficient use of SQLite in applications (especially the MaxMinion spreadsheet and similar).

Designed so that the MCP (and agents using it) can:

• Apply and evolve a spreadsheet-centric schema (cells + deps + FTS + undo + named ranges)
• Generate and execute smart queries (recursive CTEs for dependents, FTS search, "render stream" queries for UI grids)
• Advise on :memory: + ATTACH patterns, triggers for sync, user-defined functions registration strategy
• Provide CSV <-> SQLite bulk in the optimal way (using .import or virtual tables)
• Help with static build / amalgamation considerations when embedding sqlite in V binaries

All development of this MCP happens in its own Fossil repository (this checkout) so it can be published/extracted independently.

Bootstrap (follow exact fossil-scm-mcp pattern)

# (already done for this repo)
fossil open ...
./bootstrap.sh
.venv/bin/python -m sqlite_mcp
# or from MaxMinion checkout:
#   mcp/run-sqlite-mcp.sh

See also the MaxMinion project's mcp/MCP-INTEGRATION.md and mcp/example-mcp-config.json for how this MCP is wired into the spreadsheet development workflow.

Key Resources

• sqlite://best-practices
• sqlite://spreadsheet-schema
• sqlite://recalc-cte-patterns
• sqlite://gui-stream-queries

Tools (high level)

• sqlite_apply_spreadsheet_schema
• sqlite_exec (with safety)
• sqlite_get_affected_cells (CTE)
• sqlite_build_fts
• sqlite_csv_import
• sqlite_memory_engine_tips

See server.py for current impl.

Integration with MaxMinion / V

The V code in MaxMinion uses the patterns documented here (not necessarily calling the MCP at runtime; the MCP is an agent dev tool + reference impl).

Fossil Rules

Strict. See parent FOSSIL-FOR-AGENTS.md . All changes via fossil commit + tech notes where appropriate.