A self-hostable AI research workspace for grounded chat, paper study, scientific skills, and research execution.
%20%7C%2025%20Current-339933?logo=node.js&logoColor=white){kind=icon}
English · 简体中文 · 日本語 · Français · Deutsch
Documentation · Quick Start · Community
InnoClaw turns server-side folders into AI-native workspaces where you can chat over your own files, study papers, and run experiments — all in one place.
Instead of juggling separate tools for literature, notes, code, and automation, you open a folder, sync it, and work: cited answers over real documents, structured paper reviews, reusable scientific skills, and a path from reading to remote execution.
- Researchers who read papers, run experiments, and want cited AI answers grounded in their own files
- ML / AI engineers who need a workspace for code, data, and agent-assisted execution on remote clusters
- Lab teams who want a shared, self-hosted research hub instead of scattered SaaS tools
- Self-hosters who want full control over their data, models, and infrastructure
Requires Node.js 24+ (package.json is the source of truth). If you use nvm: nvm install && nvm use.
git clone https://github.com/SpectrAI-Initiative/InnoClaw.git
cd InnoClaw
npm install
cp .env.example .env.local # then edit: set WORKSPACE_ROOTS and at least one API key
mkdir -p ./data
npx drizzle-kit migrate
npm run dev # open http://localhost:3000After the UI opens: Settings → configure a model provider → open a workspace → click Sync → start chatting.
Security: InnoClaw includes shell execution and remote job submission capabilities. See SECURITY.md for deployment hardening and trust boundary documentation.
Environment variables, upgrade flow, and advanced setup
Set WORKSPACE_ROOTS in .env.local to one or more absolute paths where your research folders live:
WORKSPACE_ROOTS=/absolute/path/to/workspaces
OPENAI_API_KEY=sk-...WORKSPACE_ROOTSdirectories must already exist before startupnpx drizzle-kit migratecreates or upgrades the SQLite schema at./data/innoclaw.db- If the repo lives on NFS/CIFS, set
DATABASE_URLandNEXT_BUILD_DIRto local disk paths
Upgrading:
git fetch --tags
git checkout vX.Y.Z # or: git pull (if tracking main)
npm install
npx drizzle-kit migrate
npm run buildCheck CHANGELOG.md before every upgrade. Compare .env.local against .env.example for new variables.
For OS-specific prerequisites and production deployment: see Installation and Deployment.
git clone https://github.com/SpectrAI-Initiative/InnoClaw.git && cd InnoClaw
cp .env.production.example .env.production.local # edit: set API key + WORKSPACE_ROOTS
docker compose up -d # open http://localhost:3000See the full Docker Deployment Guide for volumes, reverse proxy, backups, and upgrades.
InnoClaw supports three primary workflows. Pick the one that matches what you need today — you can always explore the others later.
Search literature across ArXiv, PubMed, bioRxiv, and Semantic Scholar. Summarize papers, run structured multi-role discussions (moderator, skeptic, librarian, reproducer, scribe), and generate research ideation from what you read.
Start here: open Paper Study in any workspace.
Open a server folder as a persistent workspace. Chat over your files with RAG-backed citations. Browse, edit, and sync files. Use the agent panel to run multi-step tasks with tool calling. Import reusable scientific skills across domains like drug discovery, genomics, and protein science.
Start here: create a workspace, click Sync, and ask a question.
Go from code inspection to job submission and result analysis. Review repositories with agent assistance, gate high-risk steps with approval checkpoints, submit jobs through Shell, Slurm, or rjob, and monitor execution across clusters.
Start here: open Deep Research in a workspace with remote profiles configured.
- Docker Deployment Support: Added Dockerfile, docker-compose.yml, and full Docker deployment guide for self-hosted production setups
- 200+ New Built-in Skills: Expanded skill library with bioinformatics, cheminformatics, genomics, physics, and drug discovery pipelines
- Skill Creator Framework: New meta-skill with evaluation, benchmarking, and validation tooling for building and testing custom skills
- Text-to-CAD Skill: New agent skill that converts natural language descriptions into 3D CAD models (STL/STEP) using CadQuery, with automatic environment setup
- Workspace Image Picker: New dialog UI in the agent panel for browsing and selecting images from the workspace to attach to conversations
Show earlier updates
- Pasted Image Support: Users can now paste images directly into the chat input for multimodal AI conversations
- Deep Research Role Studio: New Role Studio panel lets users configure and manage custom researcher roles in the deep research workflow
- Expanded Paper Search Sources: Added BioRxiv, PubMed, and PubChem as searchable paper sources in Paper Study
- Dynamic Model Discovery: Agent panel now auto-fetches available models from each configured AI provider, merging live results with built-in model lists
- Per-Model Base URL Routing: Chinese AI providers (shlab, qwen, moonshot, deepseek, minimax, zhipu) now support per-model
<PROVIDER>_<MODEL>_BASE_URLenv vars for flexible endpoint routing - Runtime Tool-Calling Override: Tool support can now be toggled per provider via
<PROVIDER>_TOOLS_ENABLED=true/falsewithout code changes
- Node.js Runtime Update: InnoClaw now targets Node.js 24+ and is verified against both Node.js 24 LTS and the latest Node.js 25 current release. CI and local version hints have been updated accordingly.
- Multimodal LLM Support: Paper Study and agent workflows now support both standard LLMs and multimodal LLMs (mLLM), selectable per-context in settings and the model selector
- GitHub Skills Import Preview: New pre-import preview workflow lets users browse, review, and selectively import skills from GitHub repositories before committing changes
- Obsidian Note Export: Generate structured, Obsidian-compatible paper notes with rich YAML frontmatter, figures, and wikilinks directly from the paper study panel
- Per-Task Model Selector: New model selector UI component lets users override the default AI model for individual paper study tasks (summary, roast, notes, etc.)
- Note Discussion View: New full-page discussion view for paper notes, enabling threaded AI-assisted conversations around generated note content
- Remote HPC/SLURM Execution: Deep research sessions can now run on remote clusters via SSH, supporting rjob, rlaunch, and SLURM schedulers with file staging and job lifecycle management
- Kubernetes Cluster Config UI: New settings panel for runtime configuration of K8s contexts, PVC bindings, and container images across multi-cluster deployments without restarting
- Remote Profile Binding: Deep research sessions can be bound to pre-configured SSH/remote compute profiles, enabling reproducible distributed research workflows
- Deep Research Module: Full AI-driven scientific research pipeline with multi-phase orchestration, reviewer deliberation, execution planning, and workflow graph UI
- Execution Pipeline: Automated experiment execution system with Slurm job submission, dataset management, preprocessing, and remote executor support
| Feature | What it enables |
|---|---|
| Workspace Management | Map server folders into persistent AI workspaces |
| File Browser | Browse, upload, create, edit, preview, and sync files |
| RAG Chat | Ask grounded questions over indexed files with citations |
| Paper Study | Search, summarize, and inspect papers from ArXiv, PubMed, bioRxiv, and more |
| Discussion Mode | Run structured multi-role paper discussions |
| Research Ideation | Generate new directions and cross-disciplinary ideas |
| Skills System | Import reusable scientific and workflow skills |
| Deep Research | AI-driven multi-phase research with workflow graph and role-based execution |
| Research Execution | Orchestrate remote experiment loops with monitoring and approval gates |
| Multi-Agent Sessions | Keep separate execution contexts across tabs and projects |
| Multi-LLM Support | Use OpenAI, Anthropic, Gemini, and compatible endpoints |
| Layer | Role |
|---|---|
| Workspace | Files, notes, session context, and project state |
| Knowledge | RAG index over synced files for grounded answers |
| Paper Workbench | Literature search, summary, discussion, and ideation |
| Skills | Reusable domain workflows and tool-guided capabilities |
| Execution | Remote jobs, experiment loops, and result collection |
- Start here — Overview, Installation
- Configure and deploy — Deployment, Environment Variables, Configuration
- Use the product — Features, API Reference
- Troubleshoot and contribute — Troubleshooting, Development Guide
- Need setup or usage help? Start with the docs
- Found a bug or want a feature? Open an issue
- Want direct discussion? Join the Feishu or WeChat communities below
Scan to join our community · 扫码加入飞书/微信体验群
- License — Apache-2.0, see
LICENSE - Repository — https://github.com/SpectrAI-Initiative/InnoClaw
- Docs — https://SpectrAI-Initiative.github.io/InnoClaw/