Deploy AutoMem

AI long-term memory: AutoMem API and MCP bridge on FalkorDB and Qdrant

Deploy AutoMem

/data

Just deployed

/qdrant/data

/data

Deploy and Host AutoMem on Railway

AutoMem is an open-source (MIT) long-term memory service for AI assistants. It stores memories in a FalkorDB graph and a Qdrant vector index, so Claude, ChatGPT, Cursor and other MCP clients can recall decisions, preferences, relationships and context across sessions through one HTTP API.

About Hosting AutoMem

This template runs AutoMem 0.16.3 as four services that build from one public GitHub repo, each pinned by version and digest: the AutoMem API (public), an MCP bridge for cloud AI clients (public), FalkorDB 4.22 and Qdrant 1.20 (both private). Graph data, vectors and downloaded embedding models sit on Railway volumes, so they survive redeploys. The API token, admin token, FalkorDB password and Qdrant API key are generated for each deploy. Add a Voyage or OpenAI key for hosted embeddings, or leave both empty to use a local model. Version bumps reach you as Railway update notices.

Common Use Cases

  • Give Claude Desktop, Claude Code, Cursor or Codex a shared memory that persists between sessions and machines
  • Connect ChatGPT, Claude.ai or ElevenLabs agents to the same memory through the MCP bridge
  • Store project decisions, preferences and facts once and recall them by meaning, tags or time

Dependencies for AutoMem Hosting

  • FalkorDB (graph store, bundled, private, on a volume)
  • Qdrant (vector store, bundled, private, on a volume)
  • Optional: a Voyage AI or OpenAI API key for hosted embeddings

Deployment Dependencies

Implementation Details

After the deploy:

  1. Open the AutoMem service, copy AUTOMEM_API_TOKEN from its Variables tab and its public domain from Settings. https://YOUR-AUTOMEM-DOMAIN/health shows falkordb and qdrant as connected.
  2. Point a local MCP client at the API, for example npx @verygoodplugins/mcp-automem setup with AUTOMEM_API_URL=https://YOUR-AUTOMEM-DOMAIN and the token.
  3. For cloud clients (ChatGPT, Claude.ai), use the MCP bridge: https://YOUR-BRIDGE-DOMAIN/mcp with the header Authorization: Bearer YOUR-AUTOMEM-API-TOKEN (or ?api_token= where a client can't send headers). The bridge never uses a token of its own, so a request without a token gets 401.

Embeddings: with VOYAGE_API_KEY or OPENAI_API_KEY set, AutoMem uses that provider (Voyage first). With neither, it downloads the local bge-large-en-v1.5 model (about 1.2 GB) to the volume once and runs it in the API service, which needs about 4 GB of RAM. Keep VECTOR_SIZE=1024 unless you plan a re-embed. OPENAI_API_KEY also enables memory classification and enrichment.

Updates: the repo pins every image by digest. When we bump a version, Railway shows an update on your project, and the repo's CHANGELOG.md lists breaking changes first. FalkorDB stays on 4.x, because FalkorDB 6 can't open a 4.x volume without a migration. Admin endpoints (backup, re-embed) use ADMIN_API_TOKEN.

Why Deploy AutoMem 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 AutoMem 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

More templates in this category

View Template
Chat Chat
Chat Chat, your own unified chat and search to AI platform.

okisdev
117
View Template
Weaviate
Vector database with hybrid search and an API key-protected API

GitHub Locker
1
View Template
Möbius Personal AI OS
Self-hosted AI agent that builds and runs apps for you, with memory.

bento
1