MCP Server

What is MCP?

The Model Context Protocol (MCP) is an open standard that lets AI assistants -- like Claude Desktop, Claude Code, Cursor, and other MCP-aware clients -- connect to external data sources and tools. BRC Analytics runs a public MCP server that exposes our full catalog (organisms, assemblies, workflows), raw sequencing-data search via the European Nucleotide Archive (ENA) -- plus NCBI SRA where a local mirror is configured -- as well as read-only context resources and guided analysis prompts.

Once connected, you can ask questions like "What assemblies do you have for Plasmodium falciparum?" or "Find me Illumina RNA-seq runs for taxonomy ID 5833" in plain language, and the assistant will pull live answers from BRC Analytics instead of guessing.

This is an experimental feature -- the tool surface and connection details may change.

Connecting

The MCP endpoint is:

https://brc-analytics.org/api/v1/mcp/

The trailing slash is the canonical form; the server also accepts the URL without it.

Claude Code

Add the server to Claude Code using the CLI:

claude mcp add --transport http brc-analytics https://brc-analytics.org/api/v1/mcp/

Claude Desktop

Claude Desktop connects to remote MCP servers as custom connectors. Open Settings > Connectors, choose Add custom connector, and paste the endpoint URL:

https://brc-analytics.org/api/v1/mcp/

Alternatively, claude_desktop_config.json can reach the server through the mcp-remote bridge. That file only launches local command-based servers, so a bare url entry won't work there. Edit the config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows) and add:

{
  "mcpServers": {
    "brc-analytics": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://brc-analytics.org/api/v1/mcp/"]
    }
  }
}

Restart Claude Desktop. You should see the brc-analytics server listed in the tools panel.

Cursor and Other MCP Clients

For Cursor, add the server to ~/.cursor/mcp.json or configure it via Cursor Settings > MCP:

{
  "mcpServers": {
    "brc-analytics": {
      "url": "https://brc-analytics.org/api/v1/mcp/"
    }
  }
}

Any MCP-compliant client that supports the Streamable HTTP transport can connect using the same URL. The older SSE transport is not supported. Consult your client's documentation for the exact configuration format.

Running Locally

If you are running the BRC Analytics backend yourself (see the project README), use:

http://localhost:8000/api/v1/mcp/

(or http://localhost:8080/api/v1/mcp/ when using the Docker stack).

For Claude Desktop against a local backend, use the mcp-remote config above with the local URL.

What You Can Do With It

The server exposes tools, read-only resources, and guided prompt templates.

Catalog Tools

Explore curated BRC Analytics content:

  • search_organisms(query, limit=10) -- search organisms by scientific name, previous/other names, taxonomy ID, or taxonomic group.
  • get_organism(taxonomy_id) -- fetch a single organism record by NCBI taxonomy ID.
  • get_assemblies(taxonomy_id) -- list all genome assemblies for an organism, including ploidy, assembly level, coverage, and annotation status.
  • get_assembly_details(accession) -- get full metadata for a specific genome assembly (e.g. GCF_000005845.2).
  • list_workflow_categories() -- list all analysis workflow categories with workflow counts.
  • get_workflows_in_category(category) -- list all workflows within a category using its key (e.g. VARIANT_CALLING) or display name.
  • get_compatible_workflows(ploidies, taxonomy_id="") -- discover workflows compatible with specific ploidy levels (e.g. HAPLOID, DIPLOID; case-insensitive) and optional taxonomy filters.
  • get_workflow_details(iwc_id) -- get full workflow metadata, TRS ID, and step information by IWC ID.
  • check_compatibility(iwc_id, accession) -- check whether a workflow and genome assembly are compatible, returning ploidy, taxonomy, or annotation issues.
  • resolve_workflow_inputs(iwc_id, accession) -- resolve deterministic inputs (reference genome FASTA, gene models) and identify required user data (sequencing reads).

Sequencing Data Tools

Search repositories for raw sequencing reads to power your analyses:

ENA tools -- live search against the European Nucleotide Archive:

  • search_ena(taxonomy_id) -- find read runs by NCBI taxonomy ID (up to 50 records per call, with a has_more pagination flag).
  • search_ena_keywords(keywords) -- search runs by keywords, such as platforms (ILLUMINA, OXFORD_NANOPORE) or library strategies (RNA-Seq, WGS).

SRA mirror tools -- fast local queries (available on deployments configured with an SRA metadata mirror):

  • search_sra(organism, assay_type=None, platform=None, country=None, since=None, limit=50) -- fast, structured multi-facet search over every public SRA run as of the mirror's build.
  • sra_data_summary(organism) -- snapshot of available SRA data (run counts, top platforms, assays, BioProjects).
  • get_sra_study_runs(accession, limit=200) -- retrieve runs for an SRA study (SRP*, ERP*, DRP*) or BioProject (PRJNA*, PRJEB*, PRJDB*).

Logan search tools -- read-only access to Logan sequence searches run from the Logan search page (available on deployments configured with a Galaxy API key):

  • logan_job_status(job_id) -- check whether a Logan search has finished.
  • logan_cohort(job_id) -- whole-match-set counts and facets for a finished search.
  • logan_hits(job_id, offset=0, limit=25) -- a page of score-ranked hits with SRA metadata (limit is capped at 100).

These read the results the search page already assembled and cached for a day; if a tool reports the results as expired, open that search's results page again first.

Catalog Resources

Clients supporting MCP resources can read catalog context directly via standard URIs without consuming tool execution steps:

  • brc://catalog/summary -- JSON summary of total organisms, assemblies, workflows, and categories.
  • brc://catalog/categories -- JSON listing of all workflow categories and descriptions.
  • brc://catalog/workflows -- JSON list of all available assembly-scoped workflows and requirements, each listed once with every category it belongs to.
  • brc://catalog/organisms/{taxonomy_id} -- JSON details for a specific organism by taxonomy ID.

Guided Prompts

The server includes pre-configured workflow prompts:

  • plan_pathogen_analysis(organism, analysis_type="VARIANT_CALLING") -- step-by-step guidance for looking up an organism, matching assemblies, choosing a workflow from the given category, and locating sequencing datasets. analysis_type is a workflow category key, as listed by list_workflow_categories.

Things to Try

Once connected, try prompts like:

  • "Which BRC organisms are in the Apicomplexa group?"
  • "Show me the diploid assemblies you have, and tell me which variant-calling workflows are compatible with them."
  • "For assembly GCF_000002765.6, resolve the inputs for the variant-calling workflow. What still needs to be provided?"
  • "Find paired-end Illumina runs in ENA for taxonomy ID 5833."
  • "What's in the Transcriptomics workflow category?"

The assistant will chain tool calls together as needed -- e.g., look up an organism, list its assemblies, check compatibility, and resolve inputs in a single conversation.

Limitations

  • Read-only. The server exposes catalog, resource, and sequencing data but cannot launch workflows on Galaxy. Use the BRC Analytics web interface to start an analysis.
  • Snapshot of the catalog. Catalog data updates when new catalog builds are published, matching what powers the website.
  • External API dependencies. ENA searches query the European Nucleotide Archive directly and may fail or rate-limit during upstream outages (where enabled, the local SRA mirror keeps working when ENA does not).

Feedback

The MCP server is actively evolving. If something is missing, broken, or confusing, email help@brc-analytics.org.