mcp-limsdq

An MCP (Model Context Protocol) server that gives AI assistants lab/LIMS data-quality superpowers: validate CSV exports against a schema, infer schemas from example files, profile a dataset, and diff before/after exports for migration or ETL parity checks.

Built for the lab-informatics workflow where a bad CSV silently poisons everything downstream. Instead of eyeballing exports, ask your assistant to check the file first.

Tools

Issue codes: MISSING_REQUIRED, TYPE_MISMATCH, OUT_OF_RANGE, NOT_ALLOWED, CENSORED_NOT_ALLOWED, UNKNOWN_COLUMN.

Censored lab tokens (ND, BQL, BDL, <0.01, TNTC, ...) are first-class: schemas can allow them per column instead of failing validation.

Related MCP server: csv-analyzer-mcp

Install

pip install mcp-limsdq

Requires Python 3.10+.

Use with Claude Desktop

Add to your MCP config (claude_desktop_config.json):

{
  "mcpServers": {
    "lims-dq": {
      "command": "mcp-limsdq",
      "args": ["serve"]
    }
  }
}

Then ask: "Validate this week's assay export against our schema and flag anything odd."

Schema format

{
  "allow_extra_columns": false,
  "columns": [
    {"name": "sample_id", "dtype": "string", "required": true},
    {"name": "result", "dtype": "float", "required": true,
     "min": 0, "max": 50, "allow_censored": true},
    {"name": "status", "dtype": "string", "allowed": ["PASS", "FAIL"]},
    {"name": "run_date", "dtype": "date", "required": true}
  ]
}

Tip: point infer_schema at a known-good export to generate a starting schema, then tighten ranges and allowed lists by hand.

Local CLI (no MCP client needed)

# validate and print issues (exit 0 = valid, 1 = invalid)
mcp-limsdq check exports/assay_2026-09-21.csv schemas/assay.json

# quiet mode for CI
mcp-limsdq check exports/assay.csv schemas/assay.json --quiet --report report.json

Development

python -m venv .venv && .venv/bin/pip install -e ".[dev]"
.venv/bin/python -m pytest

The quality engine (src/mcp_limsdq/checks.py) is stdlib-only and fully decoupled from the MCP layer, so it can be reused as a plain library.

Examples

See examples/: samples.csv (clean export with censored values), samples_bad.csv (six distinct issue types), schema.json, and a migration_before.csv / migration_after.csv pair for compare.

License

MIT