gget-mcp is an MCP (Model Context Protocol) server implemented for the gget bioinformatics library, designed to provide AI assistants and agents with a standardized interface for accessing bioinformatics tools and databases. With MCP, AI can execute complex genomics queries through a structured interface.
【Purpose】 This project wraps the common functionality of the gget bioinformatics toolkit as MCP tools, enabling AI assistants to complete tasks such as genomic database queries and sequence analysis using natural language. gget itself provides efficient query capabilities for genomic databases.
【Main Capabilities】 The tools provided by the server are divided by category as follows:
- Gene information and retrieval: gget_search (find Ensembl IDs based on search terms), gget_info (retrieve detailed information for an Ensembl ID), gget_seq (retrieve nucleotide or amino acid sequences), gget_ref (retrieve reference genome information from Ensembl).
- Sequence analysis: gget_blast (perform BLAST on nucleotide or amino acid sequences), gget_blat (find the genomic location of a sequence), gget_muscle (multiple sequence alignment).
- Expression and functional analysis: gget_archs4 (retrieve gene expression data from ARCHS4), gget_enrichr (perform gene set enrichment analysis).
- Protein structure and function: gget_pdb (retrieve protein structure data from PDB), gget_alphafold (predict protein structures using AlphaFold).
- Cancer and mutation analysis: gget_cosmic (search for cancer mutations in the COSMIC database).
- Single-cell analysis: gget_cellxgene (query CellxGene single-cell RNA-seq data).
【Usage】
- Run: You can use uvx to run the gget-mcp server directly without cloning the repository. If you encounter problems, they can usually be resolved by clearing the uv cache. HTTP mode starts a web server, which can be accessed at http://localhost:3002/mcp（文档位于 http://localhost:3002/docs）。STDIO mode is intended for MCP clients that communicate via standard input/output; SSE mode uses Server-Sent Events for real-time communication.
- Configure the client: Preconfigured JSON files are provided for different use cases. STDIO mode (recommended) is suitable for most AI clients, especially when you want to save large output files (sequences, structures, alignments) to disk rather than returning them as text. There is also an HTTP mode configuration.
- Inspect tools: To view the methods provided by the server, you can run MCP Inspector with npx (Node.js and npm may need to be installed). There is a corresponding command for STDIO mode with uvx; you can also run Inspector manually and configure it through the interface. Afterward, you can browse tools and resources in MCP Inspector (usually on port 6274; running it multiple times may change the port).
- Integrate with AI systems: Point an AI client (such as Cursor, Windsurf, ClaudeDesktop, VS Code with Copilot, etc.) to the appropriate configuration file in the repository.
- If you have already cloned the repository, you can run the server with uv.
The repository also provides a step-by-step Windows guide for Claude Desktop + gget_mcp, covering Google Drive integration, including installing Google Drive for Desktop, configuring offline access, installing uv and Node.js, editing claude_desktop_config.json to add the gget_mcp server configuration, and enabling and verifying it in the Extensions and Developer tabs of Claude Desktop. Ultimately, retrieved gene sequences, protein data, and more can be saved directly to a Google Drive offline folder and synchronized automatically.
【Applicable Scenarios】 It can be used for natural language-driven bioinformatics analysis, for example: analyzing the TP53 gene and finding protein sequences and cancer mutations; predicting the 3D structure of a given protein sequence (such as BLAST identifying it as green fluorescent protein GFP, with AlphaFold predicting the classic β-barrel structure with 90%+ confidence); performing pathway enrichment analysis on DNA repair-related genes (TP53, BRCA1, BRCA2, ATM) (such as DNA Damage Response, Homologous Recombination, Cell Cycle Checkpoints, p53 Signaling). Batch operations such as searching for all apoptosis-related genes and generating a summary table, or cross-database cross-referencing (obtaining COL1A1 information from Ensembl and UniProt) are also listed as usage examples.
[Limitations and Safety] Safety features include: input validation (comprehensive parameter validation for all gget functions), error handling (robust error handling with informative messages), rate limiting (restraint in the use of external APIs and databases), data validation (type checking and data format validation). Testing and Verification: The project provides comprehensive tests, including unit tests and integration tests for network-dependent operations. The evaluation test (test_judge.py) uses a large language model to evaluate the performance of AI agents using the gget tool. It has been automatically excluded from CI/CD to avoid costs, requires an API key (GEMINI_API_KEY or similar), may incur fees from the LLM API provider, and is only used for local development and manual evaluation. GitHub Actions workflows: CI (.github/workflows/ci.yml) runs basic linting and quick tests on every push/PR; Tests (.github/workflows/test.yml) performs comprehensive testing across multiple operating systems and Python versions, including quick tests for each platform, integration tests (which may require network access), slow tests (such as BLAST operations, run only on Ubuntu + Python 3.11), code coverage reports, and security checks using bandit and safety; Publish (.github/workflows/publish.yml) publishes to PyPI when tags are pushed. Note: There is currently no Swagger/OpenAPI interface, and accessing the server directly in a browser will not provide much useful information; to explore available tools and capabilities, you should use MCP Inspector or connect through an MCP client to view them. Using MCP Inspector via npx requires Node.js and npm to be installed, and using nvm to manage Node.js versions is recommended. In addition, gget needs to retrieve data from online databases, so a stable network connection is required.
This guide was updated on October 13, 2025.
