Open-source AI memory platform that turns documents, code and conversations into a self-hosted knowledge graph for AI agents.
ModernRelay/omnigraph
Object-storage native graph database in Rust with branching, a typed ontology, and vector, full-text and graph retrieval over open Lance data.
- Stars
- 1,535
- Forks
- 234
- Open issues
- 135
- Last commit
- 7 hours ago
What is it?
Omnigraph is a graph database that stores its data directly on object storage such as S3-compatible stores, local disk, or Azure in the open Lance format. It is aimed at multiple AI agents sharing one typed model of a domain, retrieving and writing context, and coordinating work against durable state.
Schemas are declared in .pg files that define node and edge types, keys and constraints, and the engine enforces them on every write. Queries can combine graph traversal, vector search, full-text search and Reciprocal Rank Fusion in one runtime. Branches let an agent or task write changes in isolation, then merge them three-way into main, with time travel across the whole graph.
Deployments are described as code in a cluster.yaml file, with plan and apply commands in a Terraform-style workflow. Cedar policies are enforced server-side on every mutation. The project provides a CLI, an HTTP server, a TypeScript SDK, and an MCP server for LLM hosts.
Key features
- 01Typed schemas (.pg) with keys and constraints enforced on every write
- 02Combined graph traversal, vector search, full-text search and rank fusion in one query
- 03Git-style branches with three-way merge and time travel
- 04Terraform-style cluster config with plan and apply
- 05Cedar policy enforcement on the server for every mutation
- 06Data stored in open, versioned Lance format on S3-compatible storage
Use cases
- Giving multiple AI agents a shared, typed knowledge graph to read and write
- Storing agent memory with a branch per agent or task, merged after review
- Building a dev graph of issues and dependencies that coding agents can query
- Managing versioned experiment data for ML training and evaluation
Quick start
- 1
Add the Homebrew tap
brew tap ModernRelay/tap - 2
Install the omnigraph CLI with Homebrew
brew install ModernRelay/tap/omnigraph - 3
Follow the quickstart guide in the docs to set up a cluster
Things to know
- Azure Blob storage support is a qualification preview and requires a checked-in admission wrapper.
- The Python SDK is not yet available (coming soon).
Topics
FAQ
- What is omnigraph?
- Object-storage native graph database in Rust with branching, a typed ontology, and vector, full-text and graph retrieval over open Lance data. Omnigraph is a graph database that stores its data directly on object storage such as S3-compatible stores, local disk, or Azure in the open Lance format. It is aimed at multiple AI agents sharing one typed model of a domain, retrieving and writing context, and coordinating work against durable state.
- Is omnigraph free to use? What license does it have?
- omnigraph is open source under the MIT license. Check the license terms before using it in commercial products.
- Is omnigraph still maintained?
- Actively maintained. The last commit to omnigraph was on Oct 21, 2026.
- What can I use omnigraph for?
- Giving multiple AI agents a shared, typed knowledge graph to read and write Storing agent memory with a branch per agent or task, merged after review Building a dev graph of issues and dependencies that coding agents can query Managing versioned experiment data for ML training and evaluation
- How do I install omnigraph?
- Add the Homebrew tap: brew tap ModernRelay/tap Install the omnigraph CLI with Homebrew: brew install ModernRelay/tap/omnigraph Follow the quickstart guide in the docs to set up a cluster
- Where can I get omnigraph?
- The source code and installation instructions are on GitHub: https://github.com/ModernRelay/omnigraph
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A graph-native memory system for AI agents that stores conversations, builds knowledge graphs and records reasoning, backed by Neo4j.
A graph database that stores adjacency as sparse matrices and queries with OpenCypher, aimed at knowledge graphs for LLM and GraphRAG workloads.
An open-source graph database in Rust that stores graph, vector and full-text data together on object storage, queryable from Rust, TypeScript, Go and Python.
RustFS is an S3-compatible, distributed object storage system written in Rust, with Apache 2.0 licensing and a web console.
An MCP server that indexes code, docs and infrastructure into a local knowledge graph so AI coding agents can query structure instead of re-reading files.