Build context your agents can traverse
OpenGraph AI converts heterogeneous source material into queryable knowledge graphs. It preserves entities, relationships, and source evidence so an agent can follow context across multiple steps instead of searching an unstructured collection of similar passages.
Current release
opengraph-imageis the first public implementation. It builds a graph from a folder of images, then exposes that graph through a CLI and five MCP tools.
Get started
| Goal | Documentation |
|---|---|
| Build and query your first graph | Quickstart |
| Connect Claude Code or Claude Desktop | Connect Claude with MCP |
| Understand entities and relationships | Context graphs |
| Review the complete tool surface | MCP tools |
Core workflow
source images → extraction → entities + relationships → graph.json
↓
CLI / MCP → grounded answer
- Build a graph from source images.
- Inspect the extracted entities, relationships, and evidence.
- Query the graph from the CLI or an MCP-compatible agent.
- Trace each result back to its source files.
OpenGraph AI is open source under the Apache-2.0 license. The current public release focuses on images; the graph contract is designed to support additional modalities as the project develops.