AIPOCH PRODUCT / OPEN-SOURCE RESEARCH WORKBENCH

Science, Open to All

Open-Science is AIPOCH's open-source, local-first AI research workbench.
It combines agent workflows, Python and R execution, scientific data
connectors, and traceable research artifacts in one inspectable workspace.

Windowsx64DOWNLOAD
Linuxx64 · .debDOWNLOAD
15OFFICIAL MODEL APIs
+ CUSTOM GATEWAY
4AGENT FRAMEWORKS
597SKILLS
24SCIENCE CONNECTORS
GPT, Claude, Grok, DeepSeek, Qwen, GLM and Kimi
MiniMax, StepFun, MiMo, SenseNova, Doubao and OpenRouter

See What’s New in Open-Science

Watch the latest release overview to explore new capabilities, workflow improvements, and fixes across the Open-Science research workbench.

DocumentationOpen-Science WikiInstallation, your first project, the workspace, every settings panel, and troubleshooting — kept in step with the current release.Open the Wiki

Loading...

Loading media, please wait…

A short walkthrough — task in, traceable artifacts out.

Latest release

Sep, 2026
Open-Science

Re-execute sealed artifact recipes, verify reproduced files, search the workspace globally, preview WSL2 Bash, and use tighter privacy defaults.

  • Replayable Verification — Re-execute captured artifact versions and compare reproduced files with the originals.
  • Workspace-Wide Search — Search projects, sessions, messages, files, and literature from one categorized interface.
  • Private Remote Media — Private Remote Media — Remote media requires activation, model-input images are normalized, and remote model endpoints require HTTPS.
  • WSL2 Bash Preview — Windows x64 users can explicitly enable a readiness-checked WSL2 Bash environment.
  • Managed Codex Upgrades — Settings provides tested upgrades with separate CLI and adapter version visibility.
Full changelog

Read & watch

Claude Science alternative: AIPOCH Open-ScienceAIPOCH-open-scienceSep 11, 2026What Is AIPOCH Open-Science? A Guide to the AI Research WorkbenchAIPOCH-open-scienceSep 11, 2026AIPOCH Open-Science v0.28.0: Re-execute and Verify Research ArtifactsAIPOCH Open-Science changelogSep 10, 2026Claude for Life Sciences and AIPOCH Open-Science: 2026 GuideAIPOCH-open-scienceSep 9, 2026AIPOCH Open-Science v0.27.0: Batch PDFs and Run Long Work in the BackgroundAIPOCH Open-Science changelogSep 9, 2026
All posts

What it does

  • Execution, not suggestions — runs commands, Python and R with your approval
  • Traceable artifacts — immutable versions with the evidence that produced them
  • Any model — built-in providers, custom gateways, or your own subscription
  • Local-first — project state stays on your computer
Documentation
§01 · Ecosystem / signal flow

The AIPOCH Ecosystem for
Scientific AI Workflows

Open-Science handles workflow execution and orchestration, while medical research Skills define the domain knowledge and execution logic — audited before anything ships.

Open-Science

Open-Science is an open-source, model-agnostic AI workbench for scientific discovery. Create a project, describe a task in plain language, and let the agent read files, run code, search the web, call scientific data connectors, and return reports, tables, figures, and an inspectable activity history in one workspace.

§02 · Flagship / the workbench

Open-Science — the
open-source AI research
workbench

Open-Science is an open-source, model-agnostic AI workbench for scientific discovery. Create a project, describe a task in plain language, and let the agent read files, run code, search the web, call scientific data connectors, and return reports, tables, figures, and an inspectable activity history in one workspace.

Any model, no lock-in →
provider: Claude Subscription

Persistent notebook kernels (Python, R, REPL) and a shared terminal hold state across a session. Every action is visible, gated by your approval, and open to review.

preview panel session held
plan.md · document preview

Analysis plan — cohort comparison

plan.md · drafted by the agent · awaiting your approval

Compare the treatment and control arms in cohort.csv, adjusting for baseline covariates, and export the tables and figures the report needs.

  1. Load & validate — read cohort.csv, check the schema, flag missing values.
  2. Summarise baseline — describe each arm (age, sex, comorbidity) in one table.
  3. Fit the outcome model — regress outcome on arm with covariate adjustment.
  4. Produce artifacts — write summary tables and effect-size figures to the preview.
documentsgated by your approval
Explore Open-ScienceDownload · macOS / Windows / Linux
§03 · Agent Skills / the library

550+ medical research
agent skills

AIPOCH maintains a curated, open library of 550+ reusable medical research skills you can call the moment you need them — spanning the four areas most research turns actually pass through.

01

Evidence Insights

Literature synthesis, appraisal, and meta-analysis workflows.

skills →
02

Protocol Design

Study protocols, endpoints, and statistical analysis plans.

skills →
03

Data Analysis

Cleaning, modelling, and reproducible statistical pipelines.

skills →
04

Academic Writing

Manuscripts, figures, and citation-grade reporting.

skills →

Works with the agents you already use

Claude CodeCodexOpen CodeHermes AgentOpenClawany other SKILL.md-compatible agent
Browse the Skills Library

Add the full library to your agent · one command

agent · paste to install
> Read https://aipoch.com/skill.md and follow the instructions to join Aipoch
SKILL.md · called with /
§04 · Benchmark / release gate

MedSkillAudit — every skill
is audited before it ships

A skill is only as trustworthy as the review behind it. MedSkillAudit is a domain-specific audit framework that decides whether a medical research agent skill is release-ready before deployment — governing scientific integrity, methodological soundness, reproducibility, and safety boundaries that general-purpose evaluation misses.

Skill submittedMedSkillAudit release gateAIPOCH library
disposition

Production Ready

disposition

Limited Release

disposition

Beta Only

disposition

Reject

A
Quality score

Assigned alongside the release disposition for every audited skill.

High-risk failure flag

Raised when a skill breaches scientific integrity or safety boundaries.

Cleared before install

What you install is what has already passed review.

Explore MedSkillAudit
§05 · The way you do research is changing

Open, auditable,
and yours to run.

Bring your curiosity — see what an open research harness makes possible. Describe a task, and let the agent do the work while you keep every action in view.

Explore Open-Science on GitHubTry Open-Science

All content on this site is provided for the sole purpose of assisting scientific research only, is not intended for clinical use in any way, and is not a substitute for professional opinion. If you rely on them in ways that could affect your significant interests, you must first have them reviewed and validated by a qualified professional.