A curated list of AI agent frameworks, platforms, tools, and resources for building autonomous AI systems in 2026.
Unlike other lists, this one is opinionated β every entry is categorized by maturity, use case, and whether it's actually maintained. Dead projects are removed monthly.
- Agent Frameworks
- Agent Orchestration
- Coding Agents
- Research Agents
- Browser Agents
- Voice & Multimodal Agents
- Memory & State
- Evaluation & Testing
- Deployment & Infrastructure
- Learning Resources
General-purpose frameworks for building AI agents.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| LangChain | 137k+ | Python/JS | Most popular LLM framework. Chains, agents, RAG, tools. | β Active |
| LangGraph | 32k+ | Python/JS | Stateful, multi-actor agent graphs built on LangChain. | β Active |
| CrewAI | 51k+ | Python | Role-based multi-agent orchestration. Define agents with roles/goals. | β Active |
| AutoGen | 58k+ | Python | Microsoft's multi-agent conversation framework. | β Active |
| Semantic Kernel | 27k+ | C#/Python/Java | Microsoft's SDK for integrating LLMs into apps. | β Active |
| Haystack | 25k+ | Python | End-to-end NLP/LLM framework by deepset. Pipelines-first. | β Active |
| Agno | 40k+ | Python | Lightweight agent framework. Fast, multi-modal, multi-agent. | β Active |
| Swarm | 21k+ | Python | OpenAI's lightweight multi-agent orchestration. Educational. | β Active |
| Atomic Agents | 5k+ | Python | Modular, composable agents with schema-driven I/O. | β Active |
| Pydantic AI | 17k+ | Python | Type-safe agent framework by Pydantic team. | β Active |
Tools for coordinating multiple agents working together.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| MetaGPT | 68k+ | Python | Multi-agent framework that assigns roles (PM, Architect, Engineer). | β Active |
| ChatDev | 33k+ | Python | Simulates a software company with AI agent employees. | β Active |
| AutoGPT | 184k+ | Python | Autonomous GPT-4 experiment. Pioneer of agent space. | |
| BabyAGI | 22k+ | Python | Task-driven autonomous agent. Simple but influential. | |
| Camel | 17k+ | Python | Communicative agents for multi-agent cooperation. | β Active |
| Agency Swarm | 4k+ | Python | Agent communication framework using OpenAI Assistants API. | β Active |
AI agents that write, review, and debug code.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| Aider | 45k+ | Python | AI pair programming in your terminal. Git-native. | β Active |
| Codex CLI | 84k+ | Rust | OpenAI's terminal coding agent. | β Active |
| Gemini CLI | 104k+ | TypeScript | Google's terminal AI agent. | β Active |
| Cline | 62k+ | TypeScript | Autonomous coding agent for VS Code. | β Active |
| Continue | 33k+ | TypeScript | Open-source AI code assistant for VS Code / JetBrains. | β Active |
| Goose | 45k+ | Rust | Block's autonomous coding agent. MCP-native. | β Active |
| SWE-agent | 19k+ | Python | Princeton's agent that resolves GitHub issues autonomously. | β Active |
| OpenHands | 74k+ | Python | AI agents for software development (formerly OpenDevin). | β Active |
Agents that search, synthesize, and analyze information.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| GPT Researcher | 27k+ | Python | Autonomous research agent. Multi-source, generates reports. | β Active |
| STORM | 28k+ | Python | Stanford's research agent. Writes Wikipedia-style articles. | β Active |
| Khoj | 34k+ | Python | Personal AI for search and research across your data. | β Active |
| Perplexica | 34k+ | TypeScript | Open-source Perplexity alternative. AI search engine. | β Active |
AI agents that navigate and interact with web browsers.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| Browser Use | 95k+ | Python | Make AI agents interact with any website. | β Active |
| Playwright MCP | 32k+ | TypeScript | Browser automation via Model Context Protocol. | β Active |
| Stagehand | 22k+ | TypeScript | AI web browsing framework by Browserbase. | β Active |
| Skyvern | 21k+ | Python | Browser automation using LLMs and computer vision. | β Active |
| LaVague | 6k+ | Python | Large action model for browser automation. | β Active |
Agents that process voice, images, and multiple modalities.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| LiveKit Agents | 10k+ | Python | Build real-time voice AI agents. | β Active |
| Pipecat | 12k+ | Python | Framework for voice and multimodal conversational AI. | β Active |
| Vocode | 3k+ | Python | Open-source library for building voice agents. | |
| OpenVoice | 36k+ | Python | Instant voice cloning by MIT and MyShell. | β Active |
Give agents persistent memory and context management.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| Mem0 | 56k+ | Python | Memory layer for AI agents. Persistent, contextual. | β Active |
| Letta | 22k+ | Python | Stateful LLM agents with long-term memory (formerly MemGPT). | β Active |
| Zep | 4k+ | Go | Long-term memory for AI assistants and agents. | β Active |
| Chroma | 28k+ | Python | AI-native embedding database. Most popular for RAG. | β Active |
| Qdrant | 31k+ | Rust | High-performance vector database for AI. | β Active |
Test, evaluate, and monitor AI agents.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| Promptfoo | 21k+ | TypeScript | Test prompts, agents, RAGs. CI/CD integration. Used by OpenAI. | β Active |
| Langfuse | 27k+ | TypeScript | LLM observability β tracing, evals, prompt management. | β Active |
| Arize Phoenix | 9k+ | Python | ML/LLM observability and evaluation. | β Active |
| Ragas | 14k+ | Python | RAG evaluation framework. | β Active |
| DeepEval | 15k+ | Python | Unit testing for LLMs. Pytest-like interface. | β Active |
Run, deploy, and scale AI agents in production.
| Project | Stars | Language | Description | Maintained |
|---|---|---|---|---|
| LiteLLM | 47k+ | Python | Unified API for 100+ LLM providers. Cost tracking. | β Active |
| Ollama | 171k+ | Go | Run LLMs locally. Dead-simple CLI. | β Active |
| vLLM | 80k+ | Python | High-throughput LLM serving engine. | β Active |
| LocalAI | 46k+ | Go | Self-hosted OpenAI-compatible API. | β Active |
| Dify | 142k+ | Python/TS | LLM app development platform. Visual workflow builder. | β Active |
| n8n | 189k+ | TypeScript | Workflow automation with 400+ integrations. | β Active |
Guides, courses, and tutorials for building AI agents.
| Resource | Type | Description |
|---|---|---|
| LLM Engineer Handbook | Guide | Curated resources for training, serving, fine-tuning LLMs. |
| AI System Design Primer | Guide | System design for AI systems β RAG, agents, vector DBs, evals. |
| Prompt Engineering Guide | Guide | Comprehensive prompt engineering techniques and papers. |
| LangChain Cookbook | Tutorials | Official LangChain examples and recipes. |
| Anthropic Courses | Course | Official Anthropic prompt engineering and tool use courses. |
Contributions welcome! Please read the contribution guidelines first.
Rules:
- Only include actively maintained projects (commit in last 3 months)
- Include GitHub stars count (approximate is fine)
- One project per line, consistent format
- No paid-only tools without a free tier
Last updated: May 2026 | Maintained by: @vishwasvijayabaskar-code