I wanted a way to prototype an agent and have it serving requests in minutes, InitRunner is a YAML-first platform where one config file gives you a working agent with RAG, memory, and an API endpoint. apiVersion: initrunner/v1 kind: Agent metadata: name: acme-support description: Support agent for Acme Corp spec: role: You are a support agent for Acme Corp. model: provider: openai name: gpt-4o-mini ingest: sources: - ./docs/*/.md - ./knowledge/*/.pdf memory: max_memories: 1000 initrunner ingest agent.yaml initrunner run agent.yaml -i initrunner serve agent.yaml --port 3000 Three commands: ingest your docs, test interactively, serve over HTTP. The agent gets search_documents, remember, and recall tools auto-registered. No code to wire up. Point Open WebUI or any OpenAI client at localhost:3000/v1 and you have a full chat interface over your knowledge base with persistent memory across sessions. Built on PydanticAI, SQLite + sqlite-vec. No Redis, no Postgres, no infra. curl -fsSL https://initrunner.ai/install.sh | sh |