How do I deploy AgentFetch on Railway?

Deploy AgentFetch on Railway in three steps: (1) create a new Railway service from the ghcr.io/agentfetch/agentfetch:latest Docker image (Railway → New → Empty Service → Source: Docker Image); (2) set env vars in the Variables tab — AGENTFETCH_LICENSE_KEY, PORT=8080, optional AGENTFETCH_REDIS_URL pointing at a Railway Redis plugin; (3) attach a domain under Settings → Networking → Generate Domain (or Custom Domain). Total deploy time: 3-5 minutes. The container reads PORT from env via the standard Python os.getenv("PORT") pattern, so Railway's dynamic port assignment works without code changes. Memory footprint at idle is ~150MB; Railway's $5/month Hobby tier covers 500MB and is sufficient for personal use. For production with sustained 10+ req/sec workloads, bump to the Pro tier and provision a Railway Redis instance for shared cache across replicas. Healthcheck path is /health returning 200; configure under Settings → Deploy → Healthcheck Path. Once deployed, point your MCP clients at https://your-service.up.railway.app/mcp/ (trailing slash required for streamable-HTTP transport — common gotcha). Add AGENTFETCH_API_KEY as a bearer token via headers on the client side. The same pattern works on Render, Fly.io, and any container PaaS that exposes a public port and reads from env.