about
The open context layer
for reasoning agents.
We started OpenGraph AI because retrieval-augmented agents are stuck on text. Real problems live across tables, slides, recordings, and screenshots — and the answers live in the relationships between them. We're building a single, open graph layer so agents can reason across all of it.
- Tables
- Text
- Image
- Audio
- Video
relationships
- Entityappears_inSource
- Entityrelated_toEntity
- Claimsupported_bySource
agent context
Retrieved as a subgraph, not chunks.
Principles
- Open by defaultApache-2.0 core. Self-host or use the hosted version. Your graph is portable.
- Modality-agnosticIf a model can read it, we can graph it. Tables, text, image, audio, video, code.
- ComposableBring your own extractor, LLM, embeddings, or graph store. We orchestrate.
- Agent-nativeDesigned to be a single tool an agent reaches for — not a service it babysits.
beta / design partner
Talk to us
Looking for design partners, especially teams building research agents and enterprise memory.