{"slug": "omnigraph-source-control-for-context-graphs-in-rust", "title": "Omnigraph, source control for context graphs in Rust", "summary": "ModernRelay released Omnigraph, an open-source Rust-based source control system for context graphs that enables AI agents to version, branch, and merge graph data like code. The tool provides declarative schema, typed queries with vector and full-text search, and branch-based isolation for concurrent agent writes, aiming to bring software engineering discipline to knowledge management.", "body_md": "An agent reads the graph as context.\n\nThe graph is how an agent understands the business.\n\n`$ curl -fsSL https://raw.githubusercontent.com/ModernRelay/omnigraph/main/scripts/install.sh | bash`\n\nSchema, context, policies, UI and infra are declarative files: versioned like code, enforced by the engine.\n\nGit-style branch, version, and merge for data. A set of linted, typed queries defines each lens on the graph, and branches keep concurrent batch writes safe.\n\nHow it reads\n\nfind_experts ranks people three ways at once (vector similarity, BM25 full-text, and graph structure) and fuses them with rrf(). The whole lens is a file you lint, version, and review.\n\nqueries/team.gq\n\n```\nquery find_experts($topic: String) {\n    match { $p: Person }\n    return { $p.name, $p.role }\n    order { rrf(nearest($p.embedding, $topic),\n                bm25($p.bio, $topic)) }\n    limit 5\n}\n```\n\nMany agents work the same graph at once, each on its own branch, merging into main only through review.\n\nFig. A single hour on one graph: Hermes adds new data on hermes/ingest while Claude Code dedupes records on claude/dedupe.\n\nClaude Code and Hermes each fork main into their own branch.\n\ncheap fork`$ omnigraph branch create`\n\nHermes adds new nodes and edges while Claude Code dedupes existing records. Isolated branches never collide.\n\nschema + policy enforced`$ omnigraph mutate`\n\nEach branch's diff is approved by a human or another agent before it lands.\n\nhuman or agent`$ omnigraph diff`\n\nBoth land on main in review order, three-way and attributed to their agent.\n\nthree-way · audited`$ omnigraph branch merge`\n\nNo agent touches main directly. The schema rejects malformed writes at the branch, the policy decides who can merge, and the commit graph records every change and its author. Two agents, one graph, no collisions.\n\nVector, full-text, and graph traversal in one query, fused with rrf().\n\nWrites are typed and checked, rejected at the branch before they land.\n\nReaders see a consistent snapshot, never a partial write.\n\nRow-level merges with typed conflicts, not last-write-wins.\n\nEvery version stays queryable; pin a query to any snapshot.\n\nDestructive changes need explicit approval before apply runs.\n\nData, catalog, and state live in one S3-compatible bucket.\n\nEvery commit records its actor, resolved server-side.\n\nDenied and nonexistent are indistinguishable from outside.\n\nYour agents work with graph data the way engineers work with code: branched, typed, reviewed, merged, audited, and governed.\n\nAn agent reads the graph as context.\n\nThe graph is how an agent understands the business.\n\nEverything that defines it is a file.\n\nSchema, queries, policy, dashboards, deployment. Versioned like code.\n\nThe discipline code already has.\n\nEverything you trust for code now governs knowledge.\n\nAgents move at a different speed.\n\nA team migrates a schema quarterly. A fleet wants to weekly.\n\nPlan. Apply. Review.\n\nDeclarative files with a plan/apply cycle: the only form an agent can write and a human can review.\n\nAn open platform, easy to integrate — available to agents, humans, and apps through different means.\n\nTyped client for omnigraph-server: instance-per-client, typed errors, async-iterator streaming export.\n\n@modernrelay/omnigraph ↗The graph as agent tools — schema, branches, queries, mutations and ingest, bridged into LLM hosts over stdio.\n\n@modernrelay/omnigraph-mcp ↗Declarative notebooks generated from the graph — lenses an agent can write and a human can read.\n\nEdit graph content as markdown files; changes flow back as reviewed commits.\n\nOmnigraph as the context layer for Hermes agents.\n\nNot options to weigh, but guides and blueprints to follow: docs to run locally, guides for your own servers, and clone-and-go blueprints for the cloud.\n\nA folder on your machine. graphs/*.omni, zero infra.\n\nRead the docs ↗Your server plus any S3-compatible store.\n\nComing soonS3, IAM and KMS inside your own account.\n\nClone the blueprint ↗One service booting from one bucket.\n\nClone the blueprint ↗Cookbooks are runnable Omnigraph projects: schema, queries, seed data, and agent skills in one repo. Clone one, apply it, and inspect a real graph before you design your own.\n\nTrack AI and ML market signals through the SPIKE frame: signals, patterns, insights, and know-how.\n\nMonitor trials, filings, competitors, and market movement in a graph built for competitive intelligence.\n\nModel decisions, traces, actors, and artifacts so an organization becomes queryable by agents.\n\nConnect deal flow, theses, founders, portfolio signals, and follow-up work.\n\nTrack papers, claims, entities, and evidence across biotech and medical research.\n\nA personal operating graph for notes, people, commitments, routines, and recurring work.\n\nBuilt on a strong open-source foundation, aligned with the next-gen lakehouse graph.\n\nThe engine: typed, fast, memory-safe systems code.\n\nOpen columnar storage with versioning, so the graph remains readable files.\n\nThe in-memory standard for columnar data interchange.\n\nQuery execution for filtering, sorting, aggregation, and ranked retrieval.\n\nPolicy enforcement in the engine, applied before writes land.\n\nOn object storage you own — S3 or RustFS — in the open Lance format. No proprietary store, no lock-in.\n\nAgents branch the graph and propose writes in isolation; changes merge after review, like code — every change visible, reviewable and reversible.\n\nYes. On-prem runs with no outbound calls, on your hardware and network; you hold every key.\n\nThrough the TypeScript SDK against a running omnigraph-server, or via the MCP server that bridges Omnigraph into LLM hosts like Claude Desktop and Claude Code.\n\nYes — MIT-licensed. The engine, CLI and server live in the ModernRelay/omnigraph workspace.", "url": "https://wpnews.pro/news/omnigraph-source-control-for-context-graphs-in-rust", "canonical_source": "https://www.omnigraph.dev/", "published_at": "2026-06-16 13:10:21+00:00", "updated_at": "2026-06-16 13:19:10.242935+00:00", "lang": "en", "topics": ["developer-tools", "ai-agents", "ai-infrastructure", "large-language-models", "artificial-intelligence"], "entities": ["ModernRelay", "Omnigraph", "Claude Code", "Hermes", "S3"], "alternates": {"html": "https://wpnews.pro/news/omnigraph-source-control-for-context-graphs-in-rust", "markdown": "https://wpnews.pro/news/omnigraph-source-control-for-context-graphs-in-rust.md", "text": "https://wpnews.pro/news/omnigraph-source-control-for-context-graphs-in-rust.txt", "jsonld": "https://wpnews.pro/news/omnigraph-source-control-for-context-graphs-in-rust.jsonld"}}