Deploy and Host Graphiti + FalkorDB on Railway

Graphiti (by Zep) builds real-time, temporally-aware knowledge graphs for AI agents — it extracts entities, relationships, and when each fact was true from the text you feed it, so your agent gets memory that updates incrementally instead of being re-indexed from scratch. This template runs Graphiti's MCP server so any Model Context Protocol client (Claude, Cursor, Windsurf, your own app) can add episodes and search the graph over plain HTTP — backed by FalkorDB, an ultra-fast, Redis-based graph database that is far lighter on RAM than Neo4j.

About Hosting Graphiti + FalkorDB

This template provisions two linked services: the official falkordb/falkordb container as a persistent graph store (with its own volume and an optional web browser UI), and the Graphiti MCP server (zepai/knowledge-graph-mcp) exposed on /mcp/ over HTTP. Railway wires them over its private network, generates a unique database password as a sealed secret, attaches the volume, provisions TLS and a public domain, and restarts on failure. The graph store is never left open and survives redeploys. You bring only an LLM API key.

Why Deploy Graphiti + FalkorDB on Railway?

• Lower RAM, lower bill — FalkorDB is a Redis-core graph engine that idles far below a comparable Neo4j deployment, and Railway bills by GB-hour of memory.
• Real-time, time-aware memory — Graphiti adds facts incrementally and tracks their validity over time, instead of recomputing the whole graph the way classic GraphRAG does.
• MCP-native — point any MCP client at the public /mcp/ URL and your agent can write to and query a live knowledge graph immediately.
• Two clean services — inspect, scale, or back up the database independently of the app, with an optional FalkorDB browser UI for running Cypher.
• Your keys, your data — nothing is bundled; the graph persists on your own Railway volume.

Common Use Cases

• Long-term agent memory that stays current as facts change (people, projects, preferences).
• A shared knowledge graph that several agents or apps read and write over MCP.
• GraphRAG / entity-extraction pipelines that need temporal context, not just vector recall.

Dependencies for Graphiti + FalkorDB Hosting

• An LLM API key (OPENAI_API_KEY by default) for entity extraction and embeddings. Other providers (Anthropic, Gemini, Azure, Groq) are supported via Graphiti configuration.

Deployment Dependencies

• Graphiti — open-source temporal knowledge graphs by Zep (MCP server image zepai/knowledge-graph-mcp).
• FalkorDB — open-source, Redis-based graph database (official falkordb/falkordb container).

Why This Template?

It pairs Graphiti with FalkorDB instead of Neo4j for a lighter, cheaper, equally capable knowledge-graph backend — the only published Graphiti image that speaks FalkorDB, wired to the official FalkorDB container and ready to deploy in one click. Source and docs: .