Projects

Tools for transparent,reproducible and verifiable science.

Our projects turn hidden methodological choices into structured, inspectable research workflows, making it easier to compare alternatives and understand how results were produced.

Our projects

ASTRA

Agentic Schema for Transparent Research Analysis

ASTRA specificationGitHub repository

A specification language and validator for analyses. One astra.yaml declares your inputs, outputs, methodological decisions, and the prior insights from the literature that justify each option. Includes the astra CLI, machine-verified evidence quotes, and the schema everything else builds on.

lightcone-cli

Agentic execution layer

User guideGitHub repository

The lc command line, plus a Claude Code plugin (skills, agents, hooks) that turns an astra.yaml into materialized results. Recipes run through Snakemake on a Dask cluster — local laptop, SLURM allocation, or external scheduler. Every output gets a content-addressed manifest so lc verify can prove what you have on disk is what was actually computed.

Principles

  1. One spec, one truth.

    astra.yaml is the single source of truth. Every input, output, recipe, and decision lives there.

  2. Decisions, not defaults.

    Every meaningful methodological choice is a named decision with options, a default, a rationale, and (where possible) literature evidence behind it.

  3. Provenance you can prove.

    Every materialized output ships with a sidecar manifest recording the recipe, the container image, the decisions, the input chain, and content hashes. Tampering and drift are detectable, not hopeful.

  4. Agents do the typing, you do the science.

    The Claude Code plugin (/lc-new, /lc-migrate, /lc-feedback) handles the implementation grind so researchers can stay on the questions that matter.

Try it in two commands

Install the CLI, then let lc init scaffold the ASTRA specification, project structure, environment, and local Git repository.

The official tutorial follows a small analysis from this starting point through its first methodological comparison, execution, and integrity check.

Follow the Getting Started tutorial

Terminal walkthrough: $ uv tool install lightcone-cli. ✓ lightcone-cli ready. $ lc init my-analysis && cd my-analysis. ✓ my-analysis/ created