InnoClaw

Grip Science, Unleash Innovation at Scale

💡 Innovation for

🤖 💡 🧬 📚 🔬

Turn Server Folders into AI Workspaces

Chat with AI grounded in your documents via RAG, leverage 206 scientific skills, and orchestrate autonomous agents — all from a single self-hosted platform.

InnoClaw is built to democratize AI-powered scientific research. By providing a self-hosted, open-source platform that unifies document intelligence, multi-agent orchestration, and 206 domain-specific skills, it empowers researchers, labs, and institutions to accelerate discovery — without surrendering data to third-party clouds.

GitHub Documentation Quick Start SCP Skills

206 Skills

8 Domains

9 LLM Providers

§01

At a Glance

Skills

Domains

LLM Providers

Agent Tools

API Routes

Languages

§02

Core Features

A comprehensive AI research platform combining workspace management, RAG-enhanced chat, autonomous agents, and 206 scientific skills in a single self-hosted application.

Workspace + File Management

Map server-side folders as persistent workspaces. Browse, upload, edit, and preview files with tree-view navigation, Markdown rendering, and PDF support.

RAG-Enhanced Chat

AI answers grounded in your workspace documents with inline source citations. Powered by vector embeddings stored in SQLite for instant retrieval.

Smart Note Generation

Auto-generate summaries, FAQs, briefings, newsletters, timelines, and daily/weekly reports from your workspace content with one click.

AI Agent System

Autonomous agents with 36 tool-calling capabilities including bash execution, file operations, kubectl commands, article search, and scientific skill invocation.

Paper Study

Cross-source academic search spanning arXiv, HuggingFace Daily Papers, and Semantic Scholar. AI-powered query expansion and batch summarization.

Multi-Agent Sessions

Tab-based multi-session management with independent contexts, rename support, and MAX mode for automatic context summarization to prevent overflow.

Dataset Management

Download datasets from HuggingFace & ModelScope with pause/resume. Real-time progress tracking, manifest building, data preview, and workspace linking.

Research Execution

13-stage experiment workflow with SSH remote execution, Slurm job scheduling, real-time monitoring, and multi-agent collaboration across 5 AI roles.

Scheduled Automation

Cron-based task automation for daily reports, weekly summaries, git sync, and custom workflows. Auto-generate research digests on schedule.

§03

206 Scientific Skills Across 8 Domains SCP

Powered by Science Context Protocol (SCP), each skill is composable into autonomous research workflows — from drug target identification to protein structure prediction.

💊 71

Drug Discovery & Pharmacology

Target identification, ADMET prediction, virtual screening, molecular docking, pharmacovigilance.

🧬 41

Genomics & Genetic Analysis

Variant pathogenicity assessment, cancer genomics, population genetics, gene-disease associations.

🔬 38

Protein Science & Engineering

ESMFold/AlphaFold structure prediction, binding site analysis, protein-drug interaction profiling.

🧪 24

Chemistry & Molecular Science

Structure analysis, molecular fingerprints, structure-activity relationships, SMILES processing.

⚙️ 18

Physics & Engineering

Circuit analysis, thermodynamics, optics, electromagnetic field analysis, signal processing.

🔬 7

Experimental Automation

Protocol generation, PubMed literature search, experimental data processing.

🌍 5

Earth & Environmental Science

Atmospheric science calculations, oceanographic properties, seismic waveform processing.

📚 2

Literature & Mining

Scientific literature mining, biomedical web search, meta-analysis execution.

Science Context Protocol

SCP provides a standardized skill format for autonomous scientific research. Each skill includes overview, capabilities, usage examples, and API endpoint connections via the Intern-Discovery Platform.

Overview → When to Use → Capabilities → Examples

InternScience/scp

206 SKILLS ACROSS 8 DOMAINS

Multi-LLM Support 9 Providers · 26 Models

OpenAIGPT-5.2, 4.1, o3/o4

AnthropicOpus 4, Sonnet 4, Haiku

GoogleGemini 2.5/3/3.1

DeepSeekV3.2

QwenQwen3 235B, 3.5 397B

MoonshotKimi K2.5

SH-LabIntern S1, S1 Pro

MiniMaxMiniMax 2.5

ZhipuGLM-5

§04

Representative Workflows

Each workflow chains multiple skills and tools into end-to-end research automation.

Document-Grounded Research

RAGworkspacesource-citations

Scientific Computing Pipeline

agentbashkubectlvolcano-gpu

Literature Discovery & Summarization

arxivsemantic-scholarhuggingfacebatch-summary

Drug Target Analysis

scp-drug-discoveryADMETmolecular-docking

Protein Structure Prediction

scp-proteinESMFoldAlphaFoldbinding-sites

§05

Research & Discovery

Advanced multi-agent research capabilities — from structured paper discussions to automated experiment execution.

Multi-Agent Paper Discussion

A 5-role structured discussion system that simulates an academic seminar. AI agents collaborate through a 6-phase workflow to produce deep, multi-perspective analysis of any research paper.

🎓

Moderator

📚

Literature

🔍

Critic

🛠️

Replicator

✍️

Scribe

6-Phase Workflow

1Overview & Context

2Methodology Analysis

3Critical Review

4Reproducibility Check

5Cross-Perspective Debate

6Synthesis & Conclusion

Research Ideation

AI-powered brainstorming engine that generates cross-disciplinary research directions grounded in your paper collection. Discover innovation points and unexplored connections.

cross-disciplinary paper-grounded innovation-discovery

Experiment Execution

End-to-end experiment automation: SSH remote execution, Slurm/rjob scheduling, real-time job monitoring, result collection, and AI-driven analysis — all orchestrated by 5 specialized AI roles.

SSH Slurm 13-stage multi-agent

Advanced File Viewers

Built-in

📄

PDF Viewer

💻

Code Editor

📐

Markdown

🏢

3D CAD

🧬

Molecule PDB

🗃️

Get Started in Minutes

terminal

                            # Clone the repository

                            git clone https://github.com/SpectrAI-Initiative/InnoClaw.git

                            cd InnoClaw

                            # Install dependencies

                            npm install

                            # Configure environment

                            cp .env.example .env.local

                            # Edit .env.local with your API keys

                            # Initialize database & start

                            mkdir -p ./data && npx drizzle-kit migrate

                            npm run dev

                            # Open http://localhost:3000

Clone the repo, install dependencies, configure your API keys, and start the development server. InnoClaw runs on any machine with Node.js 20+.

Requirements

Node.js 20+ required

SQLite — zero-config database

Supports 9 LLM providers (OpenAI, Anthropic, Gemini, DeepSeek, Qwen, etc.)

Ready in under 3 minutes

Or via Claude Code

claude → /project:setup

§07

Integration Ecosystem

Feishu Bot

WebSocket real-time messaging with interactive card messages. Agent tool-calling directly from Feishu conversations.

Kubernetes

Submit GPU computing jobs via Volcano scheduler. Manage cluster resources directly from the Agent panel.

GitHub

Clone and pull repositories with private repo support via personal access tokens. Integrated Git operations.

HuggingFace

Browse and download datasets directly. Access Daily Papers for latest AI research. Integrated model discovery.

ModelScope

Download models and datasets from Alibaba ModelScope. Filter patterns, manifest building, and progress tracking.

Terminal

Built-in shell access for workspace command execution. Supports bash, sh, and zsh with sandboxed directory context.