10x Genomics Cloud MCP Server

Launch, monitor, and manage Cell Ranger single-cell and spatial genomics analyses on the 10x Genomics Cloud platform from inside Claude.

Healthcare & Life Sciences by 10x Genomics Bearer Token active

GitHub Docs

Overview

The 10x Genomics Cloud MCP server (txg-mcp) is the official integration between 10x Genomics Cloud Analysis and AI agents like Claude Desktop and Claude Code. It is distributed as a local Model Context Protocol Bundle (.mcpb) that wraps the TXG CLI, translating natural language requests into Cloud Analysis API calls so users can run Cell Ranger pipelines through conversation instead of clicking through the web UI.

The server exposes around 27 tools covering the full Cell Ranger workflow: creating projects, uploading FASTQ files and custom references, launching Cell Ranger count, multi, and aggr analyses, monitoring status, listing annotation models, and downloading outputs. It supports Universal 3'/5' Gene Expression and Flex Gene Expression chemistries, multimodal experiments (Cellplex, CRISPR, antibody capture), and aggregation runs. Configuration helpers (get_multi_csv_config_spec, get_aggr_csv_config_spec) let the model generate valid pipeline config CSVs from a plain English description.

The extension is published in Anthropic's Life Sciences Claude Code marketplace (10x-genomics@life-sciences) and is also available as a one-click Claude Desktop extension. It requires a 10x Genomics Cloud account and a personal access token, and is only useful when paired with FASTQ data you can upload to (or already have in) the platform.

Tools

Tool	Description
`get_tool_version`	Returns the TXG CLI version backing the MCP server.
`verify_auth`	Confirms the configured access token is valid and returns the authenticated account email.
`list_projects`	Lists all projects in the 10x Cloud account.
`create_project`	Creates a new Cloud Analysis project.
`update_project`	Updates a project's name or description.
`list_project_files`	Lists non-FASTQ files attached to a project.
`upload_project_file`	Uploads a file (e.g. config CSV) into a project.
`download_project_file`	Downloads a specific project file to a local path.
`list_fastqs`	Lists FASTQ files in a project.
`upload_fastqs`	Uploads FASTQ files into a project.
`list_libraries`	Lists FASTQ library types available for assignment.
`set_library`	Assigns a library type classification to a set of FASTQs.
`list_prebuilt_references`	Lists 10x prebuilt reference genomes.
`list_custom_references`	Lists custom reference genomes uploaded to the account.
`get_reference`	Returns metadata for a specific reference.
`update_reference`	Updates a reference's metadata.
`upload_reference`	Uploads a custom reference package.
`list_annotation_models`	Lists available cell annotation models.
`get_multi_csv_config_spec`	Returns the CSV format specification for Cell Ranger multi config files.
`get_aggr_csv_config_spec`	Returns the CSV format specification for Cell Ranger aggr config files.
`create_cellranger_count_analysis`	Launches a Cell Ranger count pipeline for single-sample gene expression data.
`create_cellranger_multi_analysis`	Launches a Cell Ranger multi pipeline for Flex assays and multimodal experiments (Cellplex, CRISPR, antibody capture).
`create_cellranger_aggr_analysis`	Launches a Cell Ranger aggr pipeline to combine multiple Cell Ranger runs.
`list_analyses`	Lists all analyses within a given project.
`get_analysis_details`	Returns status, parameters, and metadata for a specific analysis.
`list_analysis_files`	Lists output files produced by an analysis.
`download_analysis_files`	Downloads all (or selected) analysis output files to a local directory.

Setup Guide

Prerequisites

A 10x Genomics Cloud account: https://cloud.10xgenomics.com/
A personal access token from Cloud Account Settings > Security
Claude Desktop or Claude Code installed
Optional: the Filesystem desktop extension if you want Claude to upload FASTQs from a local directory

Option 1: Claude Desktop extension

Open Claude Desktop > Settings > Extensions > Browse extensions
Search for "10x Genomics" and click Install (or download the latest .mcpb file from the 10XGenomics/txg-mcp Releases page and double-click it)
Paste your 10x Cloud access token when prompted
Toggle the extension from Disabled to Enabled

Option 2: Claude Code plugin

/plugin marketplace add anthropics/life-sciences
/plugin install 10x-genomics@life-sciences

Then configure the access token:

Run /plugin, choose Manage plugins, find 10x-genomics, choose Configure
Paste your access token from https://cloud.10xgenomics.com/account/security
Restart Claude Code
Verify with /mcp

Option 3: Build from source

git clone https://github.com/10XGenomics/txg-mcp.git
cd txg-mcp
npm install
npm run build
npm run download-bin
npm run pack
npm run start YOUR_ACCESS_TOKEN

The token is persisted to a gitignored credentials.txt for subsequent runs.

Verifying the connection

In Claude, ask: What 10x cloud tools do you have access to? The model should list the available txg-mcp tools (e.g. list_projects, create_cellranger_count_analysis).

Use Cases

Spin up a new project, upload FASTQs from a local folder, and kick off a Cell Ranger count run on Universal 3' Gene Expression data without leaving the chat.
Configure complex Cell Ranger multi runs (Flex, Cellplex, CRISPR, antibody capture) by describing the experiment in plain English and letting the model draft the multi config CSV.
Batch-process many samples: upload an entire directory of FASTQs and launch parallel Cell Ranger count analyses for each sample.
Monitor running pipelines, then download metrics_summary.csv and web_summary.html for completed analyses and have the agent interpret QC metrics inline.
Aggregate multiple completed Cell Ranger runs into a single dataset via Cell Ranger aggr for downstream cross-sample comparisons.

Example Prompts

"What 10x cloud tools do you have access to, and which projects exist in my account?"
"Create a project called 'first-mcp-project' and start a Cell Ranger count analysis on the human 3' GEX FASTQs in /Users/me/data/sample1 using the GRCh38 prebuilt reference."
"Set up a Cell Ranger multi analysis named 'Flex-GEX-multi' in the existing 'Q4-screens' project for 4-plex Flex Gene Expression with a CRISPR library, and ask me for anything you need."
"Check the status of analysis 'Flex-GEX-multi', and once it finishes, download the outputs to ~/results."
"Read the metrics_summary.csv from my last count run and explain whether the cell recovery and saturation numbers look healthy."

Pros

Official server maintained by 10x Genomics and distributed through Anthropic's Life Sciences marketplace.
Comprehensive coverage of the Cell Ranger workflow: projects, FASTQs, references, annotation models, and count/multi/aggr pipelines (~27 tools).
One-click install as a Claude Desktop extension or Claude Code plugin, with built-in CSV config spec helpers that let the model generate valid multi/aggr configs.
Pairs cleanly with the Filesystem extension for true batch upload and analysis from a local directory.

Limitations

Only useful if you have a 10x Cloud account with data in it; the platform enforces concurrency limits (up to 25 concurrent analyses or 3 TB of data).
No tools for canceling in-progress analyses or deleting analyses, and iteration only works on analyses created via MCP, not via the web UI or CLI.
Claude Desktop cannot read web_summary.html files larger than 1 MB, limiting in-chat QC report inspection (Claude Code is less constrained).

Alternatives

AWS HealthOmics MCP Server for running bioinformatics workflows on AWS HealthOmics.
Benchling MCP (also in the Anthropic Life Sciences marketplace) for sample, sequence, and ELN management around wet-lab workflows.
Self-hosted Cell Ranger via the 10x Genomics CLI if you prefer running pipelines on your own HPC instead of 10x Cloud.