Skip to content

Stargazer

Stargazer is a bioinformatics pipeline framework built on Flyte v2. It provides a library of composable tasks for common bioinformatics operations and exposes them through an MCP server that any AI-powered client can consume.

How It Works

  1. Assets carry structured metadata that describes every file's role in the system — reference genomes, read pairs, alignments, variant calls
  2. Tasks are atomic operations that fetch assets, run bioinformatics tools, and produce new assets
  3. Workflows compose tasks into end-to-end pipelines (e.g., germline variant calling for a cohort)
  4. MCP server exposes tasks, workflows, and storage as tools that LLM clients can invoke

Interface

The marimo notebook is the primary user surface — the hosted app serves per-user notebooks for learning, running pipelines, and authoring (see Notebooks). For local use, each release publishes one end-user Docker image, stargazer — a Marimo notebook for running pipelines and exploring data, with Claude Code in the notebook's dropdown terminal.

The MCP server runs from a source install and connects to any MCP-compatible client (Claude Code, OpenCode, Cursor, etc.). Source contributors work natively against the repo — see Contributing.

Documentation