Genkit

If you're building AI workflows that need to be production-ready rather than notebook experiments, this is worth looking at. You get type-safe flows with Zod schemas, which means catching errors before runtime instead of debugging why your LLM input broke in production. The developer UI at localhost:4000 is genuinely useful for tracing what happened in a multi-step pipeline. It supports the models you'd expect (Gemini, OpenAI, Anthropic, Ollama) and handles the RAG plumbing with vector databases like Pinecone and pgvector. The .prompt file versioning is a small touch that matters when you're iterating on prompts across a team. Built for Firebase and Cloud Run, so deployment is straightforward if you're already in that ecosystem.

Install to Claude Code

npx -y skills add supercent-io/skills-template --skill genkit --agent claude-code