
Deploy Omnigraph
Graph database for agents: branches, typed schemas, hybrid search
Omnigraph
Just deployed
/var/lib/omnigraph
Deploy and Host Omnigraph on Railway
Omnigraph is a graph database for agents: typed schemas, Git-style branches and merges, and graph traversal, vector and full-text search in one query. This template runs omnigraph-server with its cluster on a Railway volume, behind HTTPS and a bearer token, ready for omnigraph cluster apply from your machine.
About Hosting Omnigraph
This template deploys one service from the published modernrelay/omnigraph-server image:
- Omnigraph: the HTTP server (also
/mcpfor agents) on a public HTTPS domain, with a volume at/var/lib/omnigraphthat holds the cluster, its graphs and history On first start the service creates an empty cluster whose only policy lets the actoract-adminmanage configuration and list graphs. The server then acceptsOMNIGRAPH_ADMIN_TOKENas that actor. You add graphs, schemas, stored queries and policies from your own machine with theomnigraphCLI, without a restart. After deploy:
- Install the CLI (same release as the image,
v0.16.0). - Copy
OMNIGRAPH_ADMIN_TOKENfrom the service's Variables tab and export it asOG_TOKEN. - Add a server target to
~/.omnigraph/config.yaml:
version: 1
targets: {railway: {kind: server, endpoint: "https://YOUR-RAILWAY-DOMAIN", auth: {provider: token, source: {env: OG_TOKEN}}}}
defaults: {target: railway}
- Write a cluster bundle and run
omnigraph cluster apply --config ./my-cluster. Keep a cluster policy (applies_to: [cluster]) that grantsact-adminconfig_manageandgraph_list, or you lock yourself out of later deployments. Every request without the token, or with a wrong one, gets 401.
Common Use Cases
- A shared, typed memory that several agents read and update, each on its own branch
- Knowledge graphs with hybrid retrieval (graph, vector, BM25) behind one HTTP API or MCP endpoint
- Declaring graphs, queries and access policy as code and deploying them like Terraform
Dependencies for Omnigraph Hosting
- The
omnigraphCLI on your machine to apply cluster bundles and load data - Optional: an embedding provider key if your schemas use vector search
Deployment Dependencies
- Omnigraph: https://github.com/ModernRelay/omnigraph (MIT)
- Server and HTTP contract: https://github.com/ModernRelay/omnigraph/blob/main/docs/user/operations/server.md
- Policies and actors: https://github.com/ModernRelay/omnigraph/blob/main/docs/user/operations/policy.md
Implementation Details
- The start command bootstraps the cluster once (a marker file on the volume records it), then runs
omnigraph-server --cluster /var/lib/omnigraph/cluster. Restarts reuse the applied state. OMNIGRAPH_ADMIN_TOKENis generated at deploy time and mapped toact-adminthroughOMNIGRAPH_SERVER_BEARER_TOKENS_JSON. Rotate it by changing the variable and redeploying.- Storage is the local volume, so run one replica. For S3-compatible storage, see the upstream deployment guide.
- Upgrade by changing the image tag together with your CLI; read the release notes first, since storage formats move forward only.
Why Deploy Omnigraph on Railway?
Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it. By deploying Omnigraph on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.
Template Content
Omnigraph
modernrelay/omnigraph-server:v0.16.0
