Encode Toolkit MCP Server

ENCODE Toolkit — Genomics Research Infrastructure for Claude

Search ENCODE, cross-reference 14 databases, run 7 analysis pipelines, and generate publication-ready methods — all from natural language in Claude Code.

Start from ENCODE but go everywhere: discover histone peaks, cross-reference with GWAS variants, check ClinVar pathogenicity, pull GTEx expression, analyze TF binding motifs from JASPAR, run pipelines, and generate publication-ready methods with full provenance — in one conversation.

---

Citation Notes

If you use ENCODE-Toolkit, please cite:

Alex M. Mawla. (2026). ENCODE-Toolkit: an MCP server, Claude plugin, and skills suite for ENCODE genomic data access and analysis. Zenodo. https://doi.org/10.5281/zenodo.18917511

BibTeX

@software{mawla_2026_encode_toolkit,
  author  = {Mawla, Alex M.},
  title   = {ENCODE-Toolkit: an MCP server, Claude plugin, and skills suite for ENCODE genomic data access and analysis},
  year    = {2026},
  publisher = {Zenodo},
  doi     = {10.5281/zenodo.18917511},
  url     = {https://doi.org/10.5281/zenodo.18917511}
}

---

Quick Start

Claude Code Plugin (recommended)

Start a new Claude Code session and enter:

/plugin marketplace add ammawla/encode-toolkit

/plugin install encode-toolkit

That's it. All 20 tools, 47 skills, and the MCP connector are now available.

If you only need the 20 MCP tools without the 47 workflow skills:

claude mcp add encode -- uvx encode-toolkit

npx (Node.js)

npx encode-toolkit

Or in MCP client config: { "command": "npx", "args": ["encode-toolkit"] }

pip install

pip install encode-toolkit

Then use encode-toolkit as the command in any MCP client configuration:

{
  "mcpServers": {
    "encode": {
      "command": "encode-toolkit"
    }
  }
}

Add to your claude_desktop_config.json:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "encode": {
      "command": "uvx",
      "args": ["encode-toolkit"]
    }
  }
}

No installation needed when using uvx. Just add the config and restart Claude.

Add to .vscode/mcp.json in your workspace:

{
  "mcp": {
    "servers": {
      "encode": {
        "command": "uvx",
        "args": ["encode-toolkit"]
      }
    }
  }
}

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "encode": {
      "command": "uvx",
      "args": ["encode-toolkit"]
    }
  }
}

Add to .windsurf/mcp.json:

{
  "mcpServers": {
    "encode": {
      "command": "uvx",
      "args": ["encode-toolkit"]
    }
  }
}

---

Connected Databases

ENCODE Toolkit integrates 14 databases through live API tools and guided skills.

---

What You Can Ask Claude

Search and explore

• "Find all histone ChIP-seq experiments for human pancreas tissue"
• "What ATAC-seq data is available for mouse brain?"
• "Search for RNA-seq on GM12878 cell line"
• "What histone marks have ChIP-seq data for pancreas?"

Download and track

• "Download all BED files from ENCSR133RZO to ~/data/encode"
• "Track experiment ENCSR133RZO with its publications"
• "Export citations for my tracked experiments as BibTeX"

Cross-reference databases

• "What GWAS variants overlap islet enhancers?"
• "Check ClinVar pathogenicity for rs7903146"
• "Pull GTEx expression for TCF7L2 across tissues"
• "Find JASPAR motifs for HNF4A binding sites"

Run pipelines

• "Set up a ChIP-seq pipeline for my H3K27ac experiments"
• "Run ATAC-seq analysis with ENCODE-standard QC thresholds"

Generate methods and provenance

• "Log that I created filtered_peaks.bed from ENCSR133RZO using bedtools"
• "Generate a methods section for my analysis with citations"

Experiment details

• "Show me the full details for experiment ENCSR133RZO"
• "What files are available for ENCSR133RZO?"
• "List only the BED files from ENCSR133RZO"

Bulk downloads

• "Download all FASTQs from human pancreas ChIP-seq to /data/fastqs"
• "Get the IDR thresholded peaks from these experiments"
• "Download the bigWig signal tracks for H3K27me3 in GRCh38"

Compatibility analysis

• "Are experiments ENCSR133RZO and ENCSR000AKS compatible for combined analysis?"
• "Compare these two ChIP-seq experiments"

Provenance chains

• "Show me the provenance chain for my derived files"
• "What files have I derived from ENCSR133RZO?"

---

The Problem

Using genomics databases today means:

1. Navigate web portals, click through dozens of filters
2. Manually find the right experiments and files across multiple databases
3. Write custom scripts to batch download
4. Lose track of which files came from where

With ENCODE Toolkit, just tell Claude what you need:

"Find all histone ChIP-seq data for human pancreas tissue"

Claude searches ENCODE, returns a structured table of 66 experiments with targets, replicates, and file counts. Downloads are organized by experiment with MD5 verification and full provenance tracking.

---

Available Tools (20)

Five core tools are shown below. The remaining 15 are collapsed for readability.

encode_search_experiments

Search ENCODE experiments with 20+ filters.

encode_get_experiment

Get full details for a single experiment including all files, quality metrics, and audit info.

encode_download_files

Download specific files by accession to a local directory.

encode_batch_download

Search + download in one step. Runs in preview mode by default.

encode_track_experiment

Track an experiment locally with its publications, methods, and pipeline info.

encode_list_files

List files for a specific experiment with format/type filters.

encode_search_files

Search files across all experiments with combined experiment + file filters.

encode_get_metadata

List valid filter values for any parameter.

encode_get_facets

Get live counts from ENCODE showing what data exists for given filters.

encode_get_file_info

Get detailed metadata for a single file.

encode_manage_credentials

Store, check, or clear ENCODE credentials for restricted data access.

encode_list_tracked

List all experiments in your local tracker with metadata, publication counts, and derived file counts.

encode_get_citations

Get publications for tracked experiments. Export as BibTeX or RIS for reference managers.

encode_compare_experiments

Analyze whether two experiments are compatible for combined analysis.

encode_summarize_collection

Get grouped statistics of your tracked experiment collection.

encode_log_derived_file

Log a file you created from ENCODE data for provenance tracking.

encode_get_provenance

View provenance chains from derived files back to source ENCODE data.

encode_export_data

Export tracked experiments as a table (CSV, TSV, or JSON) for Excel, R, pandas.

encode_link_reference

Link external references (PubMed, bioRxiv, ClinicalTrials, GEO) to tracked experiments.

encode_get_references

Get external references linked to tracked experiments for cross-server workflows.

---

Authentication

Most ENCODE data is public and requires no authentication. Just install and use.

For restricted/unreleased data, ask Claude: "Store my ENCODE credentials"

Credentials are encrypted using your OS keyring (macOS Keychain, Linux Secret Service, Windows Credential Locker) and never stored in plaintext. Get your access keys from your ENCODE profile.

---

Plugin Skills (47)

When installed as a Claude Code plugin, ENCODE Toolkit includes 47 literature-backed workflow skills that guide Claude through complex genomics tasks. Each analysis skill includes evidence-based quality thresholds, assay-specific metrics, and citations to primary literature.

Core Skills

Each pipeline includes a SKILL.md overview, 5-stage reference files (preprocessing through QC), a complete Nextflow DSL2 pipeline, a Dockerfile, and deployment configurations for local, SLURM, GCP, and AWS.

---

Why ENCODE Toolkit

Most genomics tools give you one thing. ENCODE Toolkit gives you the full research loop:

---

Supported Assay Types

Supported file formats: fastq bam bed bigWig bigBed tsv csv hic tagAlign bedpe pairs fasta vcf tar

---

Security and Privacy

• 100% local execution — no telemetry, no analytics, no tracking
• Credentials encrypted at rest via OS keyring with Fernet fallback
• Certificate verification enforced — no verify=False
• Rate limited to respect ENCODE's 10 req/sec policy
• MD5 verification on all downloads by default
• No data leaves your machine except queries to public APIs over HTTPS

---

Vignettes

Step-by-step walkthroughs showing real Claude sessions, including actual API output and scientific interpretation.

Every skill has a dedicated vignette in docs/skill-vignettes/ with a complete example session. Highlights:

See the full showcase for 15 detailed examples.

---

Development

git clone https://github.com/ammawla/encode-toolkit.git
cd encode-toolkit
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

Run the server locally:

encode-toolkit

Run tests:

pytest

---

Troubleshooting

• Make sure you restarted Claude Desktop after adding the config
• Verify uvx is installed: pip install uv or curl -LsSf https://astral.sh/uv/install.sh | sh

• Check your internet connection
• ENCODE API rate limit is 10 requests/sec — the server handles this automatically

• The file may require authentication. Ask Claude: "Store my ENCODE credentials"
• Or check if the file status is "released" on encodeproject.org

• Try broader filters (remove biosample_type or organ)
• Use encode_get_facets to see what data actually exists for your filters
• Use encode_get_metadata to check valid filter values

---

Author

Dr. Alex M. Mawla, PhD

License

AGPL-3.0. See LICENSE for full terms.

For commercial licensing inquiries: ammawla@ucdavis.edu