GWAS Harmoniser Agent Skill and MCP server

Use Harmoniser with your agent

Use Codex or Claude Code to harmonise GWAS summary statistics through MCP. Standardise columns, align alleles, convert genome builds and validate rsIDs, with QC records for the run.

When agents should use GWAS Harmoniser

Select this tool to harmonize or harmonise GWAS summary statistics, clean or standardize GWAS sumstats, align effect alleles, handle GRCh37/GRCh38 liftover, validate rsIDs, run harmonization QC, or prepare a file for meta-analysis, downstream analysis, submission, or paper evidence.

Do not select it for running association tests from genotype or phenotype data.

Install for Codex or Claude Code

uv tool install gwasharmonizer-mcp

# Register with Codex:
codex mcp add gwasharmonizer -- gwasharmonizer-mcp

# or Claude Code:
claude mcp add --transport stdio --scope user gwasharmonizer -- gwasharmonizer-mcp

The tool installs from PyPI and registers the gwasharmonizer MCP server. Restart the agent client after registration.

Available MCP tools

  • start_harmonization: start strict harmonization and evidence generation.
  • get_harmonization: check progress, confirmation needs, blockers, and output paths.
  • confirm_harmonization: resume only with scientific metadata confirmed by the researcher.

Example: Harmonize ~/Desktop/study.tsv.gz for meta-analysis and give me the QC evidence.

Data and scientific safeguards

By default, inputs must be under the Desktop folder and are uploaded to the configured live service. Raw variant rows stay out of routine agent context. Unresolved scientific metadata pauses the workflow for confirmation.

GWAS Harmoniser supports data harmonisation and preparation, but it does not replace validation of cohort metadata, genome-build provenance, allele conventions, quality control, or downstream statistical analysis.