Overview
ChemDataDoctor MCP is an independent TypeScript server that exposes local chemical table checks over MCP stdio. It is a separate product entry from the ChemDataDoctor web application. Source package and server metadata identify version 0.1.0; the repository has no GitHub releases or tags at the time of review. Installation from source is documented; an npm registry release has not been verified.
It reads CSV, TSV, XLS, XLSX and JSON files, or small inline row arrays. JSON supports arrays of row arrays or flat objects; nested objects and arrays are rejected. A CAS column mapping is required; name and formula columns are optional. Cleaning writes new JSON reports and cleaned CSV/XLSX copies while preserving the source and retaining duplicate rows.
The server needs no account or API key and has no hosted-service dependency at runtime. Installation downloads dependencies, including SheetJS from its official CDN. The source contains no outbound requests or telemetry. MCP clients decide which tool inputs and outputs reach their model provider; local server execution alone does not guarantee an offline assistant. Dataset calls return counts by default; optional samples contain metadata, while local artifacts contain user data.
CAS validation checks format and checksum, not registration or substance identity. Name cleanup changes whitespace and invisible characters only, without synonym resolution or systematic naming. Formula checks cover a restricted grammar and do not establish chemical plausibility or correspondence between CAS, names and formulas. Charges, explicit isotopes and polymer notation are unverifiable. Duplicate detection does not merge identities or remove rows, and does not carry state across calls.
Files are bounded to 20 MiB, 50,000 data rows, 200 columns, 2,000,000 cells across sheets and 100 sheets. Inline input has at most 201 rows including an optional header. Processing reads bounded input into memory. The configured data root is an application path safeguard, not an operating-system sandbox. First-party code is MIT licensed; dependencies retain their own licenses.
Key Features
- validate_cas: CAS format and checksum checks with optional conservative formatting fixes.
- normalize_names: whitespace and invisible-character cleanup while preserving case and chemical notation.
- validate_formulas: restricted formula grammar checks and atom counts.
- detect_duplicates: repeated CAS/text detection with indexes; records remain intact.
- clean_dataset: new local cleaned copies and reports; quality_report: inspection without file writes.
Use Cases
- Check chemical inventory spreadsheets before import.
- Inspect compound table quality through an MCP client.
- Prepare local cleaned copies with traceable original values.
How to Use
Requires Node.js 22.12+ and npm. Replace the example data path with an existing dedicated directory. Configure the MCP client to launch node with the absolute path to dist/cli.js and --data-root; the process waits for stdio requests.
git clone https://github.com/chemaiatlas/chemdatadoctor-mcp.git
cd chemdatadoctor-mcp
npm ci --ignore-scripts
npm run build
node dist/cli.js --data-root /absolute/path/to/data
Call quality_report with this input first. To write cleaned copies, call clean_dataset; set applyFixes=true to enable conservative fixes. The default output directory is .chemdatadoctor-output under the data root.
{"path":"inventory.xlsx","sheetIndex":0,"columns":{"cas":0,"name":1,"formula":2}}
GitHub · ChemDataDoctor Web · README · MIT
Sources and verification
This entry describes capabilities documented by the linked project sources. Catalog edits are not evidence of a successful run, compatibility test or scientific performance evaluation.
Read the editorial and verification policy