MCP Servers

Model Context Protocol servers that let an AI agent run the instrument, not just read about it — plant genomics, phenotyping, research-data acquisition, and the broker that schedules the GPU work.

Most of what an AI agent can reach through the Model Context Protocol is retrieval — search an index, fetch a record, summarize it. That is the easy tier, and it is crowded.

The harder tier is compute: tools that actually run the analysis, return a number that did not exist before the call, and show enough of their working that the number can be checked. These servers sit in that tier. Each one is on PyPI and speaks stdio, so it drops into Claude Code or any MCP client with a single config entry.

The servers

plant-genomics-mcp

50 tools across 23 public backends — Ensembl Plants, Phytozome, UniProt, AlphaFold DB, PDBe, InterPro, JASPAR, PANTHER, OrthoDB, AraGWAS, NCBI BLAST, Gramene, KEGG, STRING-DB, ATTED-II, BAR and more — plus cross-source synthesis, so a locus question does not become twenty browser tabs. stdio and Streamable-HTTP.

data-aggregator-mcp

Unified research-data acquisition. Search and fetch datasets across Zenodo, DataCite, NCBI omics (GEO / SRA / BioProject) and the literature (PubMed, OpenAIRE) behind one normalized model — the point being that "find me data on X" stops depending on which registry you happened to guess.

plantcv-mcp

Exposes PlantCV as a measurement instrument: plant traits and the segmentation overlay they were measured from. Returning the overlay is the whole design — a trait value with no picture of what was segmented is a number you have to take on faith, and a wrong segmentation produces a perfectly plausible one.

ldraw-mcp

Renders LDraw / LEGO models to images with real part geometry — studs, glass, tires — via headless Blender and ImportLDraw. It gives a vision-capable model eyes for brick builds, which turns "describe this model" into something it can actually check. The odd one out, and the most fun.

The thing underneath

jobd

A self-hostable, GPU-aware job broker for your own machines, with native MCP integration. Not a data server — it is what runs the work the others imply. Agents submit long jobs, the broker serializes GPU access across machines and sessions, and the job outlives the conversation that started it. It exists because "run this overnight" and "an agent starts it" are hard to hold together otherwise.

Install

All five are on PyPI:

pip install plant-genomics-mcp data-aggregator-mcp plantcv-mcp ldraw-mcp jobd

Each ships an MCP client config example in its README.

Archived and citable — concept DOIs resolve to the latest release:
plant-genomics-mcp · data-aggregator-mcp · jobd
See the publications page for full citations.