AI Dev Tools
The tools worth knowing about when building AI-powered apps — curated and opinionated.
LLM APIs
OpenAI
Most PopularThe most widely used LLM API. GPT-4o and GPT-4o mini cover most use cases. Best-in-class function calling and JSON mode.
Anthropic
Best for CodingClaude models excel at long documents, coding, and instruction following. 200k context window is the largest available.
Replicate
Run open-source models (Llama, Mistral, Stable Diffusion) via API without managing infrastructure. Pay per second.
Groq
Fastest InferenceBlazing fast inference for open-source models. If latency matters, Groq runs Llama and Mixtral at speeds nothing else matches.
SDKs & Frameworks
Vercel AI SDK
RecommendedThe easiest way to add streaming AI to a Next.js app. Provider-agnostic — works with OpenAI, Anthropic, and more out of the box.
LangChain
The most popular framework for building LLM pipelines. Great for RAG, agents, and chaining complex workflows.
LlamaIndex
Purpose-built for RAG and document Q&A. If your use case is "chat with your data", LlamaIndex handles it more cleanly than LangChain.
Vector Databases
Pinecone
Most PopularThe gold standard for vector search. Fully managed, scales to billions of vectors, and has a generous free tier to get started.
Supabase (pgvector)
If you're already using Supabase, pgvector gives you vector search without adding another service. Good enough for most apps.
Qdrant
Open-source vector database you can self-host or use managed. Great performance and a clean API.
Infrastructure & Deployment
Vercel
RecommendedThe easiest way to deploy Next.js. Free tier is generous, edge functions are fast, and the DX is unmatched.
Railway
Deploy backends, databases, and workers with almost no config. Great for when you need more than serverless functions.
Supabase
Postgres database, auth, storage, and edge functions in one. The fastest way to add a backend to an AI app.
Observability & Evals
LangSmith
Trace, debug, and evaluate LLM calls in production. Essential once you have real users hitting your AI features.
Helicone
Easy SetupLightweight proxy that logs all your LLM API calls. Easy to drop in and gives you cost tracking and latency metrics immediately.
🔧
Missing a tool?
Know something that belongs here? I'm always looking for tools worth recommending.
Suggest a tool →