Skip to content

fig-atlas

A tool that draws paper figures from a config, checks them before print, records where they came from, and manages them through a paper's life (exploration → paper figure → retirement). A field-neutral core is used together with domain packs (omics and others).

  • You adjust a figure in its config. Colours, sizes, thresholds and layout are all in a commented YAML file, so you can change them without reading Python.
  • Figures are drawn at their printed size. They are drawn 1:1 on the A4 text width, and --preflight checks for overlapping or clipped text, minimum sizes, colours that colour-blind readers cannot tell apart, and fonts.
  • Every figure keeps a record of its sources. The hashes of its input data, code and parameters are recorded, and when data change you are told which figures are now stale.
  • Procedures are commands. Creating a project, bringing in data, promoting and retiring figures are fa commands, and AI agents call the same commands as MCP tools.

Where to start

I want to Read
understand what this tool is and how to use it (for researchers, no programming needed) Guide for researchers
try it for the first time Tutorial: first figure (15 min)
make a figure from my data Data bundle contracts → fa ingest bundle
make a figure of microscope or blot images Tutorial: a microscopy figure
use it with an AI agent Connect AI agents
use it in another field Write a domain pack
assemble paper figures A figure's life
know why it is designed this way Design decisions (ADR)

fa --help lists every command. The repository's rules, for people and AI alike, are in AGENTS.md.