Deploy and Host Cognee AI Memory Platform with MCP on Railway

Cognee AI Memory Platform with MCP combines a private Cognee backend, durable Postgres + pgvector storage, and a public MCP service so AI tools can store, structure, and retrieve memory across sessions. The template runs the upstream cognee 1.5.4 images with OpenRouter-backed models and exposes the MCP server over Streamable HTTP for client integrations. It is a single-user deployment: the MCP endpoint is public and unauthenticated by design, so treat its URL like a secret.

About Hosting Cognee AI Memory Platform with MCP

Hosting this template deploys three connected services: cognee-api as the private backend, cognee-mcp as the public MCP layer, and PostgreSQL as the shared relational, graph, and vector store. Railway handles provisioning, private networking, and runtime configuration, while the template wires the MCP service to the backend over the internal network. The result is a single-user memory stack that supports ingestion, Cognify processing, search, and MCP tool access from clients like OpenCode and Claude Code. You only need to provide an OpenRouter API key; embeddings reuse it unless you set a separate EMBEDDING_API_KEY.

Common Use Cases

Give AI coding assistants persistent project memory across sessions through MCP
Build a private knowledge graph and vector memory backend for internal AI workflows
Store, structure, and retrieve context from documents, notes, and development artifacts

Dependencies for Cognee AI Memory Platform with MCP Hosting

An OpenRouter API key for LLM access
PostgreSQL with pgvector support (provisioned by the template)

Deployment Dependencies

Cognee docs: https://docs.cognee.ai/
Cognee MCP quickstart: https://docs.cognee.ai/cognee-mcp/mcp-quickstart
Railway template docs: https://docs.railway.com/templates/create
OpenCode config schema: https://opencode.ai/config.json

Implementation Details

The template deploys:

cognee-api from the upstream image cognee/cognee:1.5.4 with a /health healthcheck
cognee-mcp from the upstream cognee/cognee-mcp image in API mode, exposed publicly at /mcp
Postgres as the shared persistence layer for relational, graph, and vector data. Upstream cognee labels its Postgres graph adapter demo-grade and recommends a graph-native store such as Neo4j for heavy production workloads; this template favors one database over that.

Example OpenCode MCP configuration:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "cognee": {
      "type": "remote",
      "url": "https://YOUR-MCP-DOMAIN.up.railway.app/mcp",
      "enabled": true
    }
  }
}

Claude Code: claude mcp add --transport http cognee https://YOUR-MCP-DOMAIN.up.railway.app/mcp

Upgrading from an earlier version

Deployments created before September 2026 ran cognee 1.0.1. Before applying the update, back up the Postgres service. Your data migrates itself on first boot; the MCP service needs three variables and a public domain. Steps and details: https://github.com/RockinPaul/cognee_railway_template/blob/main/CHANGELOG.md

Upgrading from an earlier version

Why Deploy Cognee AI Memory Platform with MCP 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 Cognee AI Memory Platform with MCP 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.