Developer tooling
Building blocks for developers shipping AI features: agent frameworks, MCP integrations, vector search, memory layers, and evaluation and observability.
Frameworks and SDKs
Libraries for building, orchestrating, and observing LLM apps and multi-agent systems.
- Claude Agent SDK - Anthropic's SDK for building custom agents on the same harness that powers Claude Code, with tools, subagents, and MCP support; formerly the Claude Code SDK.
- CrewAI - Open-source Python framework for orchestrating role-playing, collaborative multi-agent systems.
- DeepAgents - LangChain's batteries-included agent harness with planning, sub-agents, a virtual filesystem, and context management for long-running tasks.
- Google ADK - Google's open-source Agent Development Kit for building, evaluating, and deploying multi-agent systems, model-agnostic and optimized for Gemini.
- LangChain - Framework and platform for building, running, and observing LLM applications.
- LlamaIndex - Data framework for connecting LLMs to private and enterprise data.
- Vercel AI SDK - TypeScript toolkit for building streaming AI applications across providers.
MCP and integrations
Platforms for building and connecting Model Context Protocol servers and agent tool integrations.
- Alpic - MCP-native cloud platform to build, deploy, monitor, and distribute MCP servers and ChatGPT apps.
- Composio - Integration layer that connects AI agents to over a thousand tools with managed authentication, tool search, and agent-friendly APIs.
Vector databases and retrieval
Embedding stores and search engines that power retrieval-augmented generation.
- Chroma - Open-source embedding database focused on developer ergonomics.
- LanceDB - Embedded and serverless vector database built on the Lance columnar format.
- Meilisearch - Open-source search engine with hybrid full-text and vector search and built-in AI-powered semantic ranking.
- Pinecone - Managed vector database used in production retrieval pipelines.
- Qdrant - Open-source vector search engine with on-disk indexing and filtering.
- Turbopuffer - Serverless vector and full-text search built directly on object storage.
- Weaviate - Open-source vector database with hybrid search and built-in modules.
Agent memory
Memory layers that give agents persistent, personalized context across sessions.
- Cognee - Open-source memory layer that builds knowledge graphs from your data for AI agents and RAG.
- Letta - Platform for stateful agents with long-term memory, formerly the MemGPT project.
- Mem0 - Open-source memory layer that gives AI agents persistent, personalized memory across sessions.
- Supermemory - Single memory API for fact extraction, user-profile building, contradiction resolution, and selective forgetting.
- Zep - Memory server for AI agents built on a temporal knowledge graph via the open-source Graphiti library.
Evaluation and observability
Tracing, testing, and monitoring to measure and debug LLM and agent behavior in production.
- Datadog LLM Observability - End-to-end tracing, monitoring, and quality and safety evaluation for LLM and agentic applications.
- Giskard - Open-source testing and evaluation framework that detects hallucinations, biases, and vulnerabilities in LLM and ML systems.
- Helicone - Open-source LLM observability proxy and analytics.
- Langfuse - Open-source tracing, evals, and prompt management for LLM apps.
- LangSmith - Tracing, evaluation, and monitoring for LangChain and plain LLM apps.
- Ragas - Open-source evaluation framework for RAG and LLM applications, with metrics for faithfulness, answer relevance, and context precision.