MOFA+ Remote Tool (MCP Server)#
TOU validation and deployment status (2026-08-16)#
MOFA+ 0.7.4, loopback discovery/call, and non-finite-input rejection passed; the 30-sample/two-view fixture returned two non-degenerate factors and variance components. Public publication, cross-user isolation, representative multi-omics accuracy, broad concurrency, and recovery remain incomplete. Authenticated private Platform import and owner testing passed on 2026-08-16; public publication and independent-caller authorization/isolation remain untested.
Operation:
run_mofa_factorsStart:
python -m tooluniverse.remote.mofa.mofa_toolEndpoint:
http://127.0.0.1:8024/mcpProvider configuration: no credential is needed for inline views, but private matrices must not be sent to shared deployments. Inputs and outputs are bounded by the published schema.
TOU check:
tu doctor --forward http://127.0.0.1:8024/mcp --jsonPrivate relay:
tu serve --share --forward http://127.0.0.1:8024/mcp --name validation-mofa --workers 1
Non-loopback binding requires TOOLUNIVERSE_API_TOKEN; otherwise keep the server on loopback. The relay requires TOOLUNIVERSE_SERVICE_KEY. Public publication and independent-caller testing were not run.
New-user check: python scripts/remote_validation/setup_skill_preflight.py --implementation mofa. Add --check-provider-env before launch, --live after launch, and --check-connect-prereqs before sharing. Live preflight checks exact MCP discovery only; it does not run or validate a model. The pinned relay SDK is not on PyPI and currently requires authorized GitHub repository access plus a configured SSH key, so a working local MCP server does not by itself prove that a new operator can share it.
The authenticated 2026-08-16 Platform matrix found all 30 private owner relays online and all 41 operations discoverable. This implementation was imported as unpublished owner draft(s), configured with a 120-second timeout, and invoked through /expert-sessions/{id}/test. Across the set, 38 unique operations passed return-schema and semantic validation; the three USPTO operations returned exact provider HTTP 403 and remain credential-blocked. Public publication, independent-caller authorization/isolation, broad saturation, and persistent supervision were not tested.
See the complete setup and verification guide.
Serves MOFA+ (Multi-Omics Factor Analysis v2; Argelaguet et al., Genome Biology 2020) — unsupervised integration of multiple omics layers measured on the same samples — as a ToolUniverse remote tool: run_mofa_factors (per-sample latent factor matrix + variance explained per view/factor).
Served remotely (not bundled) because mofapy2 carries a heavy probabilistic Bayesian inference stack on top of numpy/scipy.
Deploy#
pip install -r requirements.txt # mofapy2 + numpy
python mofa_tool.py # starts the MCP server on 127.0.0.1:8024
Inputs are inlined as a views object — {view_name: {feature: [value_per_sample]}} — with one matrix per omics view over the shared sample set. Every feature list must have one value per sample, in the same sample order. Expose remotely only behind TOOLUNIVERSE_API_TOKEN (SMCP bind guard).
Register in ToolUniverse#
Tool definition: src/tooluniverse/data/remote_tools/mofa_tools.json
(type: RemoteTool). Connect via the standard MCPAutoLoaderTool/server_url
mechanism.