Model-agnostic workspace
Use the models included with OpenScience, connect models from providers such as Claude, GPT, and Gemini, or continue using your own keys, subscriptions, and local models.
OpenScience
Freemium
OpenScienceは、科学研究のためのオープンソースAIワークベンチです。文献検索、科学データベース、コード、実験、計算、結果を一つのワークスペースにまとめ、付属モデル、外部プロバイダーキー、サブスクリプション、ローカルモデルに対応します。
OpenScienceとは？
OpenScience is an open-source workbench for agent-assisted scientific research. Its workspace combines literature, code, experiments, compute, and results so a research task can move from a question to sourced analysis and generated artifacts in one project. The product is designed to work with different model providers rather than tying the workflow to a single model.
The research agent can search literature through OpenAlex and arXiv, read full-text papers, access scientific databases through 42 connectors, and use tools and compute while working. Its skill library contains 371 procedures across 18 categories, covering areas such as biology, chemistry, physics, machine learning, data analysis, and scientific writing. The site’s example workflow compares predicted protein-stability changes with ProTherm measurements, runs Python code, and produces a plot with its result location.
OpenScienceでできること
Scientific literature and database access
Search OpenAlex and arXiv, read papers, and use 42 connectors to scientific databases across six domains, including examples such as UniProt, PDB, ChEMBL, and PubChem.
Research skill library
The agent has access to an indexed library of 371 skills in 18 categories. It loads a skill’s full instructions when a task requires a particular research procedure.
Code and compute execution
Run work in long-lived Python and R kernels or a shell. Heavier jobs can be dispatched, with approval, to an SSH host, Slurm or PBS cluster, or Modal as durable jobs.
Parallel specialist agents
A lead agent can assign bounded tasks to seven workers, including an exploration worker and specialists for machine learning, biology, physics, chemistry, data, and general tasks. Worker reports include verification requirements.
Ace managed model access
Ace provides a reviewed roster of 21 models from providers including OpenAI, Anthropic, Google, and xAI, along with managed research search through one Wallet. Users can set a monthly cap and use Ace beside their own model routes.
利用シーン
“Literature-grounded research”
Start with a scientific question, search relevant literature, read source papers, and use the resulting evidence in an analysis or report.
“Reproducible computational experiments”
Ask the agent to write and run Python or R code, inspect files, generate figures, and retain outputs such as a plotted image in the project workspace.
“Biomedical and life-science analysis”
Apply domain skills and database connectors to workflows involving biological structures, protein measurements, biomedical datasets, or related research data.
“Distributed research tasks”
Break a broad question into bounded literature, data, coding, or domain-specific tasks and run specialist workers in parallel.
“Lab or team infrastructure”
Send approved, compute-intensive jobs to a workstation, SSH host, Slurm or PBS cluster, or Modal while keeping the research task connected to its results.
よくある質問
What is OpenScience?
OpenScience is an open-source AI workbench and scientific research agent for reading papers, writing code, running experiments, accessing scientific data, and organizing results.
Do I need a separate AI subscription?
No. The site says free models are included, and users can also connect their own provider keys, subscriptions, or local models. ChatGPT Plus or Pro users can sign in with OpenAI to use an existing subscription.
Can OpenScience run code and use remote compute?
Yes. It supports Python and R kernels and a shell. Approved heavier jobs can be sent to an SSH host, Slurm or PBS cluster, or Modal as durable jobs.
What is Ace?
Ace is an optional managed service built into OpenScience. It offers a reviewed set of models and managed research search through one Wallet, with pay-per-request usage and an optional monthly spending cap.
Is OpenScience limited to browser use?
The site provides a browser workspace and also lists a download and CLI installation flow. The available material does not document the full platform differences or CLI capabilities.
クイック情報
- Category
- Open-source AI workbench for scientific research
- Primary workflow
- Literature, code, experiments, compute, and results in one workspace
- Scientific resources
- 371 skills and 42 database connectors
- Model approach
- Included models, external provider keys, subscriptions, local models, and optional Ace managed routes
- Compute environments
- Python, R, shell, SSH, Slurm, PBS, and Modal
- Product site
- openscience.sh
OpenScience のトラフィック分析
トラフィックデータは参考情報としてご利用ください。
- ドメイン評価
- 9
OpenScienceの代替品
Baserow
baserow.io
Baserowは、データベース、アプリ、ダッシュボード、自動化を構築できるオープンソースのノーコードDB基盤です。クラウドとセルフホストに対応します。
Relaticle
relaticle.com
Relaticleは、共有レコード、パイプラインとタスクの管理、組み込みのAIチャット、39種類のツールを備えたMCPサーバーを統合した、チーム向けのオープンソースCRMです。ホスト型ワークスペースを利用するか、AGPL-3.0のアプリケーションをセルフホストし、MCPとREST API経由で外部AIエージェントに接続できます。
Appwrite
appwrite.io
Build app backends and deploy web frontends with an open-source platform for auth, databases, files, and functions.
Dograh
www.dograh.com
Dograhは、本番向け音声エージェントを構築・デプロイできるオープンソースのセルフホスト型プラットフォームです。電話、音声、言語モデル、音声合成、ワークフロー、可観測性を接続できます。
Appwrite
imagine.dev
Appwriteは、バックエンドサービス、ファイルストレージ、リアルタイム機能、ホスティング型ウェブデプロイを提供するオープンソースのアプリケーションプラットフォームです。Auth、Databases、Storage、Functions、Messaging、Sitesを一か所で利用でき、開発者のアプリ構築とスケールを支援します。
Openship
openship.io
Openship is an open-source deployment platform for building and running applications from Git repositories, local folders, or templates. Teams can use its dashboard, CLI, desktop app, or MCP interface to deploy on Openship Cloud, connected servers, or their own infrastructure.