Directory

The AI builder's toolkit

Frameworks, models, platforms, and tools we use and recommend for building production AI agents and LLM applications.

20 resources

LangChain

Framework for building LLM-powered applications with chains, agents, and memory.

frameworkOpen SourcePythonJS

LangGraph

Build stateful, multi-actor agentic applications with graph-based orchestration.

frameworkOpen SourceAgentsOrchestration

LlamaIndex

Data framework for LLM applications, specializing in ingestion, indexing, and retrieval.

frameworkOpen SourceRAGPython

CrewAI

Framework for orchestrating multi-agent AI systems with role-playing agents.

frameworkOpen SourceMulti-AgentPython

AutoGen

Microsoft's framework for multi-agent conversation and task completion.

frameworkOpen SourceMulti-AgentPython

Claude (Anthropic)

Anthropic's family of AI models — Claude 3.5 Sonnet and Opus offer state-of-the-art reasoning and instruction following.

modelAPIReasoning

GPT-4o (OpenAI)

OpenAI's multimodal flagship model with vision, audio, and text capabilities.

modelAPIMultimodal

Llama 3 (Meta)

Meta's open-weight large language models, suitable for fine-tuning and local deployment.

modelOpen SourceOpen WeightFine-tuning

Gemini (Google)

Google's multimodal AI model family with strong coding and reasoning capabilities.

modelAPIMultimodal

Pinecone

Managed vector database for production-grade semantic search and RAG applications.

platformVector DBRAG

Weaviate

Open-source vector database with hybrid search, multi-tenancy, and built-in ML modules.

platformOpen SourceVector DBOpen Source

Qdrant

High-performance vector similarity search engine written in Rust, with filtering support.

platformOpen SourceVector DBRust

Vercel AI SDK

TypeScript SDK for building AI-powered applications with streaming, tool calling, and multi-provider support.

libraryOpen SourceTypeScriptStreaming

LangSmith

Observability and evaluation platform for LLM applications. Trace, debug, and test agents.

toolObservabilityEvaluation

Ollama

Run LLMs locally on your machine. Supports Llama, Mistral, Gemma, and more.

toolOpen SourceLocalOpen Source

Promptfoo

Open-source LLM testing and red-teaming tool. Evaluate prompts, compare models, catch regressions.

toolOpen SourceTestingEvaluation

Helicone

LLM observability platform with one-line integration — logging, monitoring, and caching.

toolOpen SourceObservabilityCaching

Instructor

Structured outputs from LLMs using Pydantic. Makes LLMs return validated, typed Python objects.

libraryOpen SourcePythonStructured Output

Semantic Kernel

Microsoft's SDK for integrating LLMs into .NET, Python, and Java applications.

libraryOpen SourceMicrosoft.NET

DSPy

Framework for algorithmically optimizing LM prompts and weights. Replaces manual prompt engineering.

libraryOpen SourcePrompt OptimizationPython