OpenCrab

What is OpenCrab

OpenCrab is an ontology-centered Graph RAG platform that structures documents into reusable AI knowledge packs. It turns messy knowledge into reusable ontology packs by ingesting documents, web research, and domain expertise, then extracting entities, concepts, evidence, relationships, and claims into an evidence graph. The platform uses a nine-space MetaOntology structure (Subject, Evidence, Claim, Policy, Lever, etc.) and supports querying through Graph RAG and MCP.

How to use OpenCrab

1. Ingest: Upload documents (PDF, HWP), URLs, notes, or structured research bundles. OpenCrab supports OCR, CLIP, Notion, and GitHub imports.
2. Structure: The system extracts entities, concepts, evidence, relationships, and claims, organizing them into a nine-space MetaOntology structure.
3. Store: The ontology graph is stored for retrieval, reasoning, and pack reuse.
4. Query: Query the graph from OpenCrab, MCP clients, or connected AI workspaces (Claude, GPT, custom agents).

Features of OpenCrab

• Ontology Packs: Reusable knowledge packs with manifest, nodes, edges, and evidence.
• Graph RAG: Query through graph-based retrieval augmented generation.
• MCP Support: Connect to MCP clients and tools.
• Multi-format Ingestion: PDF, OCR, CLIP, Notion, GitHub, expert notes.
• Nine-space MetaOntology: Structured ontology with Subject, Evidence, Claim, Policy, Lever, etc.
• Crab Agent & CrabSkill: Workflow, validator, and playbook for repeated execution.
• Marketplace: Publish and license ontology packs.
• Desktop & Web App: Available as webapp and desktop application.

Use Cases of OpenCrab

• Marketing and content planning
• YouTube channel strategy and scripts
• Brand, product, and audience analysis
• Research note synthesis
• Domain ontology pack creation
• Graph RAG over private knowledge

Pricing

FAQ

What is an ontology pack? An ontology pack is a reusable knowledge package containing manifest, nodes, edges, and evidence, structured from documents and research.

How does Graph RAG work? Graph RAG retrieves information from the ontology graph to provide grounded answers with sources you can audit.

Can I connect OpenCrab to AI tools? Yes, OpenCrab supports MCP (Model Context Protocol) for connecting to Claude, GPT, and custom agents.

Who is OpenCrab for? Researchers, analysts, domain experts, and AI builders who need to structure knowledge into reusable packs and query via graph RAG.