# synthetic-sciences/openscience

The open-source AI workbench for scientific research

Repository: https://github.com/synthetic-sciences/openscience
Canonical: https://ross.abutalabs.com/products/openscience
Homepage: https://openscience.sh
Language: TypeScript
License: Apache-2.0
License Family: permissive
Topics: agent, ai, bun, cli, llm, ml, open-source, research, science, ai-agent, co-scientist, ml-engineering, research-tools, scientific-computing, opensource
Last push: 2026-08-26T22:30:30+00:00

## Health v2 (maintenance only)
Score: 80/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 99, longevity 4
- inputs: {"age_days": 61, "days_push": 7, "days_rel": 7, "gap_med": 0, "n_releases_24m": 46}
- flags: young
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 3342, forks 452 (observed 2026-08-28T04:07:56.962844+00:00)

## What it is
OpenScience is an open-source AI workbench for scientific research that runs the full research loop: literature review, hypothesis formation, code writing and execution, experiments, and write-up. It provides a browser-based workspace with an adaptive research agent, 311 bundled domain skills, and direct access to major scientific databases like UniProt, PDB, and arXiv.

## Use cases
- automate literature review for a research topic
- run machine learning experiments from a natural language goal
- query scientific databases like UniProt and arXiv via an AI agent
- write and execute analysis code for biology or chemistry research
- generate research write-ups and LaTeX papers automatically
- fine-tune and evaluate models with DeepSpeed, PEFT, or TRL
- manage a research workspace with files, terminal, and session history

## When to choose
- you want an AI collaborator to run end-to-end scientific research workflows
- you need model-agnostic tooling that works with frontier, open-weight, or local models
- you work in ML, biology, physics, or chemistry and want integrated domain skills and database access
- you prefer open-source tools with an extensible plugin and MCP ecosystem

## When to avoid
- you need a fully offline tool with no account linking
- you only want a simple chatbot without code execution or experiment running
- your research domain lacks bundled skills and requires heavy customization
- you require guaranteed reproducibility of agent-driven experiments

## Facets
- artifact type: application
- maturity: active
- function: agent-framework, llm-inference, rag, machine-learning, developer-tools, cli, chat-interface
- domain: artificial-intelligence, large-language-models, data-science, developer-tools
- platform: bun, browser, cli, cross-platform
- tags: ai-workbench, scientific-research, co-scientist, literature-review, hypothesis-testing, scientific-databases, research-automation, model-agnostic, typescript-sdk, mcp, ai-agents, research, nodejs

## Member repositories
- synthetic-sciences/openscience (main) score 80

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:07:56.962844+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:41:03.736555+00:00, confidence not recorded.
  - readme: https://github.com/synthetic-sciences/openscience (fetched 2026-08-28T04:07:56.962844+00:00, sha 3232105eaa90)
  - homepage: https://openscience.sh (fetched 2026-08-29T09:34:31.418106+00:00, sha b9888cb26200)
- Data as of 2026-08-30T08:39:29.467469+00:00.