A lesson on using an AI coding assistant with MCP tool servers for an ePIC analysis: find a
dataset, read it, and fit the Λ⁰ → p π⁻ invariant-mass peak at 1.1157 GeV. Everything runs in
eic-shell, which includes the servers and the opencode assistant. No data download or
credentials are needed.
| Path | Contents |
|---|---|
episodes/ |
The six episodes |
learners/ |
Setup page and glossary |
instructors/ |
Instructor notes |
files/ |
Example assistant configs, AGENTS.md, and the lambda-fit skill |
extras/ |
The same analysis without MCP (uproot, RDataFrame, TTreeReader, PODIO) |
Inside eic-shell:
eic-mcp up
eic-mcp config opencode
opencode/mcp in opencode should list uproot, xrootd, and rucio. Then follow the episodes.
The lesson uses The Carpentries Workbench. With Docker, run make preview and then
make serve. With R, run sandpaper::serve().
Lesson text is CC-BY 4.0; code is MIT.