CAN-IMMUNE

A hosted, read-only MCP server that queries CAN-IMMUNE, a cancer neoantigen catalogue of ~4.49 million unique mutant peptides annotated with cell line, tissue, cancer type, and predicted MHC class I binding.

   
Type MCP server (remote, Streamable HTTP)
Supplier Li/Purcell Lab, Monash University
Availability GA — v1.0.0 published to the MCP Registry 2026-08-10 as io.github.sanjaysgk/can-immune, status active
Pricing Free / OSS (server code MIT); anonymous access at a light rate limit, higher quota after Google sign-in
Capabilities Read-only — parameterized, row-capped, time-bounded queries; no writes
Verified works · 2026-08-17
Security caution · 2026-08-17 — endpoint live (405/406 to browser GET), MIT server code, but new low-traffic repo over COSMIC/DepMap-derived data

How to install

  • Claude Code — remote MCP add (no local install, no API key): 
    claude mcp add --transport http canelib https://canelib.erc.monash.edu/mcp
  • Claude.ai — Settings → Connectors → Add custom connector, paste https://canelib.erc.monash.edu/mcp, then Connect.
  • Claude Desktop — Desktop has no native HTTP transport, so proxy the remote server through mcp-remote in claude_desktop_config.json: 
    {
  "mcpServers": {
    "canelib": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://canelib.erc.monash.edu/mcp"]
    }
  }
}

    (Unverified — upstream documents Claude Code and Claude.ai only; the mcp-remote entry above is the standard Desktop proxy pattern for a Streamable-HTTP server, not a snippet published by the project.)

Nothing runs locally: the server is hosted by Monash, so there is no process to start or keep alive.

What it does

Twelve read-only tools over the neoantigen catalogue:

  • Orientation — database_overview (record counts and coverage), list_tissues, list_cancer_types.
  • Generic retrieval — search and fetch (the pair ChatGPT-style connectors expect).
  • Genes — search_genes, get_gene, get_gene_mutations.
  • Cell lines — search_cell_lines, get_cell_line, top_genes_in_cell_line.
  • Immunology — get_mhc_binding, returning predicted MHC class I binding for a mutant peptide against a cell line’s HLA class I type.

Records are built from COSMIC and DepMap/CCLE mutations, with PubMed for literature context. Each mutation is represented as a 25-mer peptide centred on the mutated residue in both wild-type and mutant form, with anchor positions (2 and 9 of the relevant 9-mer window) surfaced; MHC-I binding is predicted with NetMHCpan (eluted-ligand mode) via the IEDB API. The database also reports which tissues and cell lines share an identical mutant peptide, so shared neoantigens can be prioritised over private ones.

Primary use cases: shared-neoantigen shortlisting for cancer vaccines, cell-line selection for immunopeptidomics experiments, HLA-aware triage of tumour mutations.

Notes

Anonymous use is allowed at a light rate limit. Signing in with Google (OAuth 2.1) raises the quota — academic domains (.edu, .edu.au, .ac.*) get the highest tier, other verified emails a standard tier; the project states sign-in reads only the verified email address to set the rate and stores nothing further.

Two hostnames are in circulation: the MCP endpoint and connector documentation are served from canelib.erc.monash.edu, while the browsable database UI is at canelib.erc.monash.edu.au. Use the .edu form in the install command above, exactly as published.

The server code is MIT-licensed, but the underlying content is derived from COSMIC and DepMap/CCLE — check those sources’ own terms before redistributing extracted data or using it commercially. MHC-I binding values are predictions, not measured immunogenicity; binding is necessary but not sufficient for a neoantigen to be immunogenic, so treat scores as a filter rather than a result. The repository is new (2 commits, 0 stars as of 2026-08-15) even though the hosted service and its Registry entry are live, and no database publication was locatable — the site asks that CAN-IMMUNE be cited but does not yet give a reference.

For prediction-side workflows rather than database lookup, see MHC binding prediction, MHC class II prediction, epitope prediction, and neoantigen prediction.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.