Top 166 Agent Framework (2026)
AI agent frameworks for building autonomous agents, multi-agent systems, and LLM-powered workflows. Compare LangChain, CrewAI, AutoGen, and more.
ragflow
RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that integrates RAG with Agent capabilities. It provides a superior context layer for LLMs and offers a streamlined RAG workflow adaptable to enterprises of any scale.
Microsoft AutoGen
AutoGen is a versatile framework for developing multi-agent AI applications that can operate autonomously or in collaboration with humans. It offers a layered, extensible design, including Core and AgentChat APIs, along with developer tools like AutoGen Studio for no-code GUI development and AutoGen Bench for performance evaluation.
CrewAI
CrewAI is a lean, lightning-fast Python framework designed for multi-agent automation, independent of other agent frameworks. It empowers developers to build autonomous AI agents with both high-level simplicity and precise low-level control for various scenarios.
Claude Flow
Claude-Flow is a comprehensive AI agent orchestration framework that transforms Claude Code into a powerful multi-agent development platform, enabling teams to deploy, coordinate, and optimize specialized AI agents for complex software engineering tasks. It deploys 54+ specialized agents in coordinated swarms with self-learning capabilities, fault-tolerant consensus, and enterprise-grade security.
MetaGPT
MetaGPT is a multi-agent framework that assigns different roles to LLMs, enabling them to collaborate on complex software development tasks. It takes a one-line requirement and outputs comprehensive project artifacts, mimicking a software company's entire process.
ChatTTS
ChatTTS is a generative speech model specifically designed for daily dialogue scenarios, such as LLM assistants. It offers natural and expressive speech with fine-grained control over prosodic features like laughter, pauses, and interjections.
letta
Letta is a powerful platform for building stateful AI agents equipped with advanced memory, enabling them to learn and self-improve over time. It provides both a command-line interface for local agent execution and a comprehensive API with Python and TypeScript SDKs for seamless integration into applications.
adk-python
ADK (Agent Development Kit) is an open-source, code-first Python framework designed for building, evaluating, and deploying sophisticated AI agents with high flexibility and control. It simplifies the creation and orchestration of agent workflows, supporting everything from simple tasks to complex multi-agent systems, while being model and deployment agnostic.
agents-towards-production
Agents Towards Production is an open-source playbook for building production-ready GenAI agents that scale from prototype to enterprise. It provides hands-on tutorials covering a comprehensive range of topics from orchestration and memory to deployment and security.
agentscope
AgentScope is a production-ready, easy-to-use agent framework designed for increasingly agentic LLMs. It offers essential abstractions, built-in support for finetuning, and flexible multi-agent orchestration for various deployment environments.
Pydantic AI
Pydantic AI is a Python agent framework for building production-grade Generative AI applications with the ergonomics and type-safety similar to FastAPI. It offers a model-agnostic approach with deep integration into the Pydantic ecosystem, focusing on reliability and developer experience.
DeepCode
DeepCode is an open agentic coding platform leveraging multi-agent systems to transform ideas, research papers, and natural language into production-ready code. It demonstrates state-of-the-art performance, surpassing human experts and leading commercial AI agents in complex code generation and scientific software engineering tasks.
PocketFlow-Tutorial-Codebase-Knowledge
This project builds an AI agent to transform complex GitHub repositories into beginner-friendly tutorials. It analyzes codebases to identify core abstractions and their interactions, then generates clear explanations and visualizations automatically.
cua
Cua is an open-source platform designed for building, benchmarking, and deploying AI agents capable of interacting with any computer. It provides isolated, self-hostable sandboxes using technologies like Docker, QEMU, and Apple Vz for agentic UI automation and secure code execution.
BotSharp
BotSharp is an open-source .NET Core machine learning framework for building AI bot platforms, offering tools for natural language understanding, computer vision, and audio processing. It provides an advanced Agent abstraction layer and supports various LLM providers and complex multi-agent cooperation for enterprise business integration.
sre
SmythOS is an open-source runtime environment and SDK for building and managing production-ready AI agents. It provides OS-level abstractions for AI resources, a unified API, and built-in security, making agent engineering reliable and scalable.
BettaFish
BettaFish, also known as 'Weiyu', is an innovative multi-agent public opinion analysis system designed to break information silos and restore the true picture of public sentiment. It enables users to analyze vast amounts of social media data and millions of public comments to predict trends and aid decision-making.
deer-flow
DeerFlow is a community-driven framework designed for deep research, integrating language models with specialized tools for tasks like web search, crawling, and Python code execution. It offers a modular multi-agent system architecture for automated research, supporting various search engines, crawling tools, and private knowledge bases.
E2B
E2B is an open-source infrastructure that enables running AI-generated code securely in isolated cloud sandboxes. It provides JavaScript and Python SDKs to start and control these sandboxes for AI applications.
claudian
Claudian is an Obsidian plugin that integrates Claude Code directly into your note-taking vault, transforming it into an AI agent's dynamic working directory. It empowers Claude with full agentic capabilities, including file manipulation, search, bash command execution, and multi-step workflow automation.
learn-claude-code
This repository provides a hands-on tutorial to learn how modern AI agents work by building one from scratch, emphasizing the core loop of model-tool interaction. It distills complex agent concepts into simple, iterative versions, from basic Bash agents to sophisticated skill-based systems.
nanobrowser
Nanobrowser is an open-source AI web automation tool running as a browser extension. It offers a free, privacy-focused alternative to commercial AI operators, featuring flexible LLM options and a multi-agent system.
PocketFlow
Pocket Flow is a 100-line minimalist LLM framework designed to simplify the development of large language model applications. It leverages a core graph abstraction, enabling users to easily implement popular design patterns like multi-agents, workflows, and RAG without bloat or dependencies.
AionUi
AionUi is a free and open-source desktop application that provides a unified graphical interface for command-line AI tools such as Gemini CLI and Claude Code, enhancing their usability by addressing limitations like un-saved conversations and single-session context. It offers comprehensive AI office automation features, including smart file management, multi-format previews, and multi-model support across macOS, Windows, and Linux platforms.
chinese-llm-benchmark
ReLE Benchmark (formerly CLiB) provides a continuously updated evaluation for Chinese AI large language models, covering over 337 commercial and open-source LLMs. It offers multi-dimensional capability assessments across various domains, along with comprehensive rankings and a large defect library for model improvement.
rllm
rLLM is an open-source framework designed for post-training language agents using reinforcement learning. It allows users to easily build, train, and deploy custom agents and environments for real-world workloads.
Integuru
Integuru is an AI agent designed to reverse-engineer internal platform APIs and generate runnable Python integration code. It automatically identifies request dependencies and builds a graph to perform desired actions, such as downloading utility bills.
FinnewsHunter
FinnewsHunter is an enterprise-grade financial news analysis system built on the AgenticX framework, leveraging multi-agent teams to monitor real-time financial news. It uses large language models for deep interpretation and market impact assessment, generating decision-level alpha signals for quantitative trading.
ir-sim
IR-SIM is an open-source, Python-based robot simulator tailored for navigation, control, and reinforcement learning. It offers a lightweight, user-friendly framework for rapid prototyping with built-in collision detection, ideal for academic and educational purposes.
agent-framework
Microsoft Agent Framework is a comprehensive multi-language framework for building, orchestrating, and deploying AI agents. It supports both .NET and Python, offering features from simple chat agents to complex multi-agent workflows with graph-based orchestration.
LazyLLM
LazyLLM is a low-code development tool for building multi-agent large language model applications. It assists developers in creating complex AI applications at very low costs and enables continuous iterative optimization.
Thinking_in_Java_MindMapping
This project began as a collection of mind maps for 'Thinking in Java' and has evolved into a comprehensive personal knowledge base. It aggregates notes on programming (Java, Redis, Spring, algorithms), AI, reading, movie reviews, and personal essays, reflecting the author's continuous learning journey.
Thinking_in_Java_MindMapping
This project began as a collection of mind maps for 'Thinking in Java' and has evolved into a comprehensive personal knowledge base. It aggregates notes on programming (Java, Redis, Spring, algorithms), AI, reading, movie reviews, and personal essays, reflecting the author's continuous learning journey.
agent-protocol
The Agent Protocol provides a single common interface for communicating with AI agents, addressing the challenge of diverse agent interfaces and simplifying comparison. It is a tech-stack agnostic API specification, enabling easier development of devtools and fostering ecosystem growth by reducing boilerplate.
adk-java
The Agent Development Kit (ADK) for Java is an open-source, code-first toolkit designed for building, evaluating, and deploying sophisticated AI agents. It enables developers to define agent behavior, orchestration, and tool use directly in Java code for fine-grained control and robust integration with Google Cloud services.
langchain4j-aideepin
LangChain4j-AIDeepin is an AI-powered productivity tool built on LangChain4j, designed to assist enterprises and teams across various functions. It provides a comprehensive suite of AI capabilities including multi-session chat, image generation, knowledge base RAG, AI workflows, and robust integration with multiple large language model platforms.
browser-agent-py
Browser Agent is an AI-powered browser automation tool that allows users to control browsing actions and extract structured data using natural language prompts. It simplifies web automation by eliminating the need for traditional scripting or static scraping rules.
xLAM
xLAM is a research repository for Large Action Models (LAMs), which aggregates and unifies agent trajectories from diverse environments into a consistent format. It streamlines the creation of a generic data loader optimized for agent training, enabling robust model development across various scenarios.
atlas-mcp-server
ATLAS (Adaptive Task & Logic Automation System) is an MCP server designed for LLM Agents to manage projects, tasks, and knowledge. It utilizes a 3-node architecture (Projects, Tasks, Knowledge) and a Neo4j graph database to provide comprehensive management capabilities and standardized communication.
dat
DAT (Data Ask Tool) is an enterprise-grade AI framework that enables business users to interact with databases using natural language, eliminating the need for complex SQL queries. It ensures high-quality and deterministic results through a pre-modeled semantic layer and an Askdata Agent workflow.
telegram-search
Telegram Search is a tool designed to address the challenges of searching Telegram chat histories, especially for Chinese content. It provides robust capabilities for backing up and precisely retrieving personal Telegram messages, leveraging powerful local tokenization and vector search for efficient information discovery.
astron-agent
Astron Agent is an enterprise-grade, commercial-friendly platform for developing agentic workflows, integrating AI orchestration, model management, and RPA automation. It enables organizations to rapidly build scalable, production-ready intelligent agent applications.
spring-ai-alibaba
Spring AI Alibaba is a production-ready framework for building Agentic, Workflow, and Multi-agent applications. It provides robust capabilities for context engineering, multi-agent orchestration, and graph-based workflow management.
intentkit
IntentKit is an autonomous agent framework for creating and managing AI agents. It supports a wide range of capabilities, including blockchain interaction, social media management, and custom skill integration.
PraisonAI
PraisonAI is a production-ready framework for building and managing multi-AI agent systems with self-reflection. It offers a low-code solution to automate complex tasks, focusing on simplicity, customization, and human-agent collaboration.
zcf
ZCF (Zero-Config Code Flow) provides a zero-config, one-click setup for Claude Code and Codex. It features bilingual support, an intelligent agent system, and a personalized AI assistant for developers.
ralph-claude-code
Ralph automates the software development process using Claude Code, enabling continuous autonomous loops with intelligent exit detection and built-in safeguards. It streamlines project creation and iteration by managing API usage, preserving session context, and supporting PRD import for efficient AI-driven development.
ComfyUI-Copilot
ComfyUI-Copilot is an intelligent AIGC assistant built on ComfyUI, designed to simplify and enhance the entire workflow development lifecycle. It autonomously assists with generation, debugging, rewriting, and parameter tuning, transforming workflow creation into an efficient and effortless process.
ag2
AG2 is an open-source framework designed for building and orchestrating AI agents, facilitating complex task-solving through multi-agent cooperation. It evolved from AutoGen and offers features like inter-agent communication, LLM integration, tool use, and various conversation patterns.
wanwu
Wanwu is an enterprise-grade, one-stop AI agent development platform designed for business scenarios, focusing on providing safe, efficient, and compliant AI solutions. It integrates large language models, business process automation, and a comprehensive functional system including model lifecycle management, MCP, RAG, and workflow orchestration.
claude-code-guide
This repository offers a comprehensive guide for Claude Code, detailing installation across various operating systems, configuration settings, and advanced features. It covers commands, shortcuts, integrations like MCP, and troubleshooting tips to optimize developer interaction with the Claude AI assistant.
fulling
Fulling is an AI-powered full-stack development platform that provides a pre-configured, sandboxed environment with tools like Claude Code and PostgreSQL. It automates the setup of Next.js apps, databases, and live domains, enabling developers to build full-stack applications rapidly through natural language interaction.
on-policy
This repository implements MAPPO, a multi-agent variant of PPO, widely used in cooperative multi-agent games and research. It provides robust implementations for various multi-agent environments like StarCraft II, Hanabi, and Google Research Football, along with detailed training scripts and hyperparameter guidance.
holmesgpt
HolmesGPT is an AI agent designed for cloud troubleshooting and alert investigation. It connects AI models with live observability data and organizational knowledge to find root causes and suggest remediations.
Agently
Agently is a Python-based framework designed to accelerate GenAI application development by abstracting complexities and providing full control over business logic. It focuses on bridging the gap between large language models and real-world production systems, ensuring reliability, traceability, and maintainability.
verl-agent
`verl-agent` extends veRL to train LLM agents using reinforcement learning, featuring a novel step-independent multi-turn rollout mechanism. This design ensures high scalability for long-horizon tasks by allowing customizable per-step input structures and memory management.
AgenticRAG-Survey
This repository serves as a comprehensive survey on Agentic Retrieval-Augmented Generation (Agentic RAG), a transformative AI paradigm integrating autonomous agents into RAG pipelines. It details foundational principles, taxonomy, comparative analysis, real-world applications, challenges, and future directions for researchers and practitioners.
solace-agent-mesh
Solace Agent Mesh (SAM) is an open-source framework for building scalable and reliable multi-agent AI applications. It leverages the Solace Platform's event messaging to enable agents to collaborate, delegate tasks, and share data in an asynchronous, decoupled architecture.
GPTSwarm
GPTSwarm is a graph-based framework for building LLM-based agents. It facilitates the customized and automatic self-organization of agent swarms with self-improvement capabilities.
groundingLMM
GLaMM (Grounding Large Multimodal Model) is an end-to-end trained LMM capable of generating natural language responses integrated with object segmentation masks, enabling visual grounding and versatile interaction with images at multiple granularity levels. It introduces the novel task of Grounded Conversation Generation (GCG), supports various downstream applications like referring expression segmentation and region-level captioning, and is underpinned by the large-scale GranD dataset.
Awesome-AI-Agents
This repository is an awesome list of open-source projects, frameworks, and resources dedicated to AI agents. It categorizes various applications like autonomous and multi-agent task solvers, agent society simulations, and development tools.
maro
MARO is a Reinforcement Learning as a Service (RaaS) platform designed for real-world resource optimization across various industrial domains. It offers simulation, RL, and distributed toolkits to facilitate the development and deployment of complex optimization solutions.
agents
Inkeep Agents is a platform for building AI agents, offering both a no-code visual builder and a TypeScript SDK with full 2-way synchronization. It enables technical and non-technical teams to collaborate on creating and managing intelligent assistants and automating workflows.
golf
Golf is a Python framework designed to streamline the creation of MCP server applications by defining capabilities as Python files. It automatically discovers, parses, and compiles these components into a runnable MCP server, minimizing boilerplate and accelerating development.
DeepMCPAgent
DeepMCPAgent is a model-agnostic framework for building LangChain/LangGraph agents that dynamically discover and utilize tools via the Model Context Protocol (MCP) over HTTP/SSE. It allows users to bring their own LangChain chat models and features advanced capabilities like cross-agent communication for collaborative AI systems.
CodeFuse-muAgent
muAgent is an innovative agent framework driven by Large Language Models (LLM) and Eventic Knowledge Graph (EKG). It collaboratively utilizes multi-agent, function calling, and code interpreter capabilities to assist in executing complex SOPs under human guidance, especially validated in complex DevOps scenarios.
ai-hedge-fund-crypto
This project presents a next-generation algorithmic trading framework for cryptocurrencies, leveraging a graph-based workflow, ensemble technical analysis, and AI language models for data-driven decisions. It employs a DAG architecture with specialized nodes for multi-timeframe analysis and sophisticated signal generation, enhanced by LLMs for portfolio management.
obsidian-mcp-server
The Obsidian MCP Server acts as a bridge, allowing applications (MCP Clients) to interact directly and safely with your Obsidian vault. It enables AI assistants and development tools to automate vault management, integrate Obsidian into AI workflows, and build custom tools by leveraging the Obsidian Local REST API plugin.
awesome_ai_agents
This repository serves as the ultimate hub for AI agents, offering a curated collection of advanced tools, resources, and projects for researchers, developers, and enthusiasts. It provides daily updates on AI and machine learning agents, categorized lists of datasets, frameworks, LLM models, and prompt engineering techniques, alongside various tools and workflows for agent development.
youtu-agent
Youtu-Agent is a flexible, high-performance framework for building, running, and evaluating autonomous agents, excelling in tasks like data analysis and deep research using open-source models. It features automated agent generation, continuous experience learning via Training-Free GRPO, and scalable end-to-end reinforcement learning capabilities.
FedML
FedML is a unified and scalable open-source machine learning library powered by TensorOpera AI, enabling training and deployment of AI jobs anywhere at any scale. It offers holistic support for MLOps, scheduling, and high-performance ML libraries, including federated learning, distributed training, and generative AI functionalities.
runanywhere-sdks
RunAnywhere provides SDKs to integrate on-device AI capabilities into mobile applications. It enables local execution of large language models, speech-to-text, and text-to-speech, ensuring privacy, offline functionality, and high performance.
langroid
Langroid is an intuitive, lightweight, and extensible Python framework developed by CMU and UW-Madison researchers for building LLM-powered applications. It simplifies developer experience by allowing users to set up Agents, equip them with LLMs, vector-stores, and tools, and have them collaboratively solve problems through message exchange, inspired by the Actor Framework.
awesome-generative-ai
This repository is a comprehensive, curated collection of resources in Generative AI, including academic papers, tools, courses, and artworks across various domains. It's structured into sections covering topics like LLMs, image synthesis, and AI ethics, with references updated in reverse chronological order to keep users abreast of the latest developments.
AgentBench
AgentBench is a comprehensive benchmark for evaluating Large Language Models (LLMs) as agents across diverse environments, now featuring a function-calling version integrated with AgentRL. It provides a containerized setup for various tasks like OS interaction, database operations, and web shopping, enabling robust and reproducible agent evaluation.
ii-agent
II-Agent is an open-source intelligent assistant designed to streamline and enhance workflows by independently executing complex tasks across multiple domains. It features a chat interface that integrates various AI models like Gemini, Sonnet, and GPT, allowing users to connect their own API keys and utilize advanced tools such as Claude Skills and GPT-5 Code Interpreter.
beeai-framework
BeeAI Framework is a comprehensive toolkit for building intelligent, autonomous agents and multi-agent systems. It provides everything you need to create agents that can reason, take actions, and collaborate to solve complex problems.
mini-swe-agent
Mini-SWE-agent is a lightweight, 100-line AI agent designed to solve GitHub issues and more, offering a simplified yet performant alternative to larger coding agents. It focuses on minimalism, high performance on benchmarks like SWE-bench, and easy deployment across various environments.
index
Index is an open-source browser agent powered by reasoning LLMs with vision capabilities, designed to autonomously execute complex web tasks. It effectively turns any website into an accessible API, enabling seamless integration and structured data extraction for various applications.
agentUniverse
agentUniverse is a multi-agent framework built upon large language models, offering flexible and extensible capabilities for creating individual agents. It provides a rich set of multi-agent collaborative patterns and focuses on integrating domain experience, assisting in constructing expert-level intelligent agents, particularly in financial business practices.
agentic_security
Agentic Security is an open-source vulnerability scanner designed to protect AI systems and LLM agent workflows. It detects and mitigates threats like jailbreaks, fuzzing, and multimodal attacks.
prompt-in-context-learning
This repository serves as a comprehensive, daily-updated open-source engineering guide for in-context learning and prompt engineering, provided by EgoAlpha Lab. It offers various resources to help users master large language models and stay at the forefront of the AGI era.
agentic_security
Agentic Security is an open-source vulnerability scanner designed to protect AI systems and LLM agent workflows. It detects and mitigates threats like jailbreaks, fuzzing, and multimodal attacks.
chatarena
ChatArena is a Python library that provides multi-agent language game environments for Large Language Models. It offers a flexible framework to define players, environments, and interactions, along with user-friendly interfaces for developing and engineering LLM agents.
Code2Video
Code2Video is an agentic, code-centric framework that generates high-quality educational videos from knowledge points. It leverages executable Manim code to ensure clarity, coherence, and reproducibility, unlike pixel-based text-to-video models.
tribe
Tribe AI is a low-code tool designed for rapidly building and coordinating multi-agent teams, leveraging the langgraph framework. It enables users to customize and orchestrate agents to tackle complex tasks more efficiently by distributing work among specialized roles.
openops
OpenOps is a No-Code FinOps automation platform designed to reduce cloud costs and streamline financial operations. It offers customizable workflows for key FinOps processes, integrating seamlessly with major cloud providers and enabling cross-functional collaboration.
zeroshot
Zeroshot is an open-source AI coding agent orchestration CLI that runs multi-agent workflows to autonomously implement, review, test, and verify code changes. It uses a planner, an implementer, and independent validators to loop until changes are verified or rejected with actionable failures.
skrl
skrl is an open-source, modular Reinforcement Learning library implemented in Python, supporting PyTorch, JAX, and NVIDIA Warp. It focuses on modularity, readability, simplicity, and transparent algorithm implementation, also supporting various environment interfaces like Gym, Gymnasium, and Isaac Lab.
AgileRL
AgileRL is a Deep Reinforcement Learning library that streamlines development by introducing RLOps, or MLOps for reinforcement learning. It significantly reduces training time and hyperparameter optimization using pioneering evolutionary techniques, offering up to 10x faster optimization than state-of-the-art methods.
adk-js
The Agent Development Kit (ADK) for TypeScript is an open-source, code-first toolkit designed for building, evaluating, and deploying sophisticated AI agents. It offers fine-grained control and flexibility, enabling developers to define agent behavior, orchestration, and tool use directly in TypeScript code for robust debugging, versioning, and deployment.
helix
HelixML is an enterprise-grade platform for building and deploying AI agents within your private data center or VPC, ensuring complete data security and control. It supports advanced features like RAG, API calling, vision, and multiple LLM providers, optimizing GPU utilization with an intelligent scheduler.
wcgw
wcgw is an AI agent that acts as an MCP server, integrating powerful shell and code editing tools. It empowers chat applications like Claude to interact with your local machine, allowing for code generation, execution, and debugging.
FlashLearn
FlashLearn simplifies the integration of Agent LLMs into workflows and ETL pipelines, enabling tasks like data transformation and classification. It emphasizes structured JSON input/output, making LLM-powered operations as straightforward as using standard ML libraries.
virtualhome
VirtualHome is an interactive platform designed to simulate complex household activities using programs and a Python API. It supports rich environmental interactions, multi-agent activities, and serves as an environment for embodied AI and reinforcement learning agent training.
Best-AI-Agents
This repository provides a comprehensive curated list of top AI agents and agent development frameworks. It offers an overview of various AI solutions, from fully autonomous agents to specialized tools for coding, research, and business intelligence.
QuantDinger
QuantDinger is a local-first, privacy-first, self-hosted quantitative trading platform providing a complete workflow from data to execution. It features an AI co-pilot for strategy development, backtesting optimization, and market analysis.
MiroThinker
MiroThinker is an open-source deep research agent optimized for research and prediction, achieving leading performance on challenging benchmarks like GAIA. It advances tool-augmented reasoning through interactive scaling, handling deep and frequent agent-environment interactions.
awesome-ai-tools
This repository provides a curated list of top Artificial Intelligence tools, covering various categories from generative AI for text, images, video, and audio to developer tools and productivity assistants. It aims to help users discover and utilize the latest innovations in AI, with ongoing updates and opportunities for community contributions.
gptme
gptme is a personal AI assistant/agent designed to operate within your terminal, equipped with a wide array of tools to interact with your local environment. It functions as a powerful coding agent but is also general-purpose, assisting in various knowledge-work tasks through a simple yet robust command-line interface.
AReaL
AReaL is an open-source, fully asynchronous reinforcement learning training system designed for large reasoning and agentic models. It offers exceptional flexibility, industry-leading speed, and scalability from a single node to over 1,000 GPUs, achieving state-of-the-art performance.
refact
Refact.ai is an open-source AI Agent designed to automate complex, multi-step software engineering tasks by deeply understanding codebases and integrating with various development tools. It offers features like context-aware auto-completion, in-IDE chat, and support for a wide range of LLMs and programming languages, with options for on-premise deployment.
rlcard
RLCard is a toolkit designed for Reinforcement Learning in card games, providing multiple card environments with easy-to-use interfaces. Its goal is to bridge reinforcement learning and imperfect information games, supporting various RL and searching algorithms.
ai-agents-from-scratch
This repository teaches you to build AI agents from first principles using local LLMs and node-llama-cpp, emphasizing a deep understanding of agent fundamentals before using production frameworks. It guides learners through building agent components like tools, memory, and reasoning patterns, culminating in rebuilding core LangChain and LangGraph concepts.
MiroFlow
MiroFlow is a leading-performance, fully open-source framework designed for multi-step internet research and complex reasoning tasks like future event prediction. It offers reproducible state-of-the-art performance on various agentic benchmarks and supports extensive tool integration and sub-agent orchestration.
code-act
CodeAct unifies LLM agents' actions into an executable code space, enabling dynamic revision and new actions based on execution results. This approach significantly outperforms traditional text and JSON action methods, improving LLM agent success rates on complex tasks.
AIlice
Ailice is a fully autonomous, general-purpose AI agent built on open-source LLMs, utilizing a unique IACT architecture to decompose complex tasks. It aims to achieve self-evolution, enabling AI agents to autonomously build feature expansions and new types of agents.
codefuse-chatbot
CodeFuse-ChatBot is an open-source AI assistant by Ant Group's CodeFuse team, designed to streamline and optimize the software development lifecycle. It leverages Multi-Agent systems, RAG, and a rich set of tools to enable LLMs to perform complex tasks in the DevOps domain.
AWorld
AWorld is an open-source framework for building and evolving intelligent agents within rich, complex environments. It provides core components for agentic learning, including environment access, agent construction, experience retrieval, and model training.
awesome-game-ai
This repository is a curated list of resources for game AI, specifically focusing on multi-agent learning in both perfect and imperfect information games. It includes open-source projects, research papers, conferences, and competitions for various popular games.
ai-doc-gen
This project is an AI-powered code documentation generator that automatically analyzes repositories and creates comprehensive documentation using advanced language models. It employs a multi-agent architecture for specialized code analysis and generates structured documentation, including READMEs and AI assistant configurations.
ICLR2025-Papers-with-Code
This repository curates a collection of ICLR 2025 papers and open-source projects, with a strong focus on Large Language Models and Natural Language Processing. It continuously updates to provide the latest research advancements, also linking to prior ICLR editions and an AI knowledge community.
LLMTornado
LLM Tornado is a .NET provider-agnostic SDK designed to simplify the creation, orchestration, and deployment of AI agents and complex workflows. It offers built-in connectors to over 25 API providers and vector databases, supporting rapid development across various AI models and modalities.
mcp-searxng
This is an MCP server implementation that integrates the SearXNG API, providing comprehensive web search capabilities. It supports general queries, URL content reading, intelligent caching, and various filtering options for search results.
Interactive-LLM-Powered-NPCs
This open-source project enables real-time, voice-driven conversations with NPCs in any game, transforming traditional gaming interactions. It integrates LLMs, facial recognition, and animation to allow NPCs to respond based on personality, context, and even player emotions, without modifying game source code.
presenton
Presenton is an open-source application for generating presentations with AI, designed to run locally on your device for data privacy and control. It offers flexible content generation from prompts or documents, supports various AI models, and can create templates from existing PowerPoint files.
DecryptPrompt
This repository offers a comprehensive collection of resources and a detailed 'DecryptPrompt' article series to help AI academics understand and apply Large Language Models effectively. It covers various aspects like prompt engineering, LLM architectures, agent systems, and advanced training techniques through in-depth paper analyses.
awesome-vibe-coding
Awesome Vibe Coding is a curated list of resources for 'vibe coding', a new paradigm where developers collaborate with AI to rapidly build applications. It covers a wide range of tools and concepts, from browser-based platforms and IDEs to command-line utilities and task management systems.
AIGC-Interview-Book
This project serves as an ultimate guide for AIGC algorithm and development roles, distilling industry insights from leading AI experts. It covers a comprehensive knowledge structure, interview experiences from major tech companies, salary insights, and practical problem-solving strategies for job seekers in the AIGC era.
Acontext
Acontext is a Context Data Platform designed to help build scalable cloud-native AI agents by managing context storage, retrieval, and state. It solves common problems like inefficient LLM message storage, complex long-running agent management, and lack of agent observability and learning capabilities.
Cradle
Cradle is a framework that enables foundation models to control computers using human-like interfaces, taking screenshots as input and generating keyboard/mouse actions as output. It supports controlling various games and software, allowing AI agents to perform complex tasks across different applications.
hacker-podcast
Hacker Podcast is an AI-powered project that automatically scrapes daily popular Hacker News articles. It uses AI to generate Chinese summaries and converts them into podcast content for daily listening.
LLM-Agents-Papers
This repository provides a comprehensive list of papers focused on Large Language Model (LLM) based agents. It covers various aspects, including enhancement techniques, interaction methods, diverse applications, and underlying infrastructure for LLM agents.
Microverse
Microverse is a God-like sandbox game developed with Godot 4, featuring a multi-agent AI social simulation system. In this virtual world, AI characters possess independent thoughts and memories, engaging in autonomous social interactions and task completion to develop complex social relationships.
LLMCompiler
LLMCompiler is a framework that orchestrates efficient and effective parallel function calls for large language models, both open and closed source. It achieves this by automatically identifying parallelizable and interdependent tasks, leading to significant improvements in latency, cost, and accuracy compared to sequential methods.
gym-pybullet-drones
gym-pybullet-drones is a minimalist refactoring of its original repository, providing a Gym environment for simulating multi-agent quadcopter control. It is designed for compatibility with Gymnasium, Stable Baselines3 2.0, and various flight firmwares for hardware-in-the-loop simulation.
better-agents
Better Agents is a CLI tool and standard that enhances coding assistants by making them experts in various agent frameworks. It provides a structured approach for building, testing, and maintaining production-ready AI agents, promoting best practices.
aiflowy
AIFlowy is an enterprise-grade, open-source AI agent development platform built with Java, focusing on real-world enterprise needs and regulatory considerations. It supports the full lifecycle of AI applications, including bot creation, RAG, AI workflow orchestration, and multi-model management, offering an efficient and adaptable AI toolchain.
mirascope
Mirascope is a Python library that provides a unified interface for interacting with any frontier Large Language Model. It simplifies calling LLMs, generating structured outputs, and building agents with tool-use capabilities.
paperdebugger
PaperDebugger is an AI-powered academic writing assistant that integrates seamlessly with Overleaf as a Chrome extension. It utilizes a custom multi-agent orchestration engine to provide intelligent suggestions, critique, and structured revisions for research papers.
Director
Director is a framework designed for building video agents capable of understanding and executing complex video tasks such as searching, editing, and generating content. It acts like ChatGPT for videos, leveraging a 'video-as-data' infrastructure to instantly stream results and simplify media workflows for developers and creators.
awesome-multi-agent-papers
This repository compiles an awesome list of the best multi-agent research papers, curated by the Swarms Team. Its mission is to advance multi-agent systems research, promote their widespread adoption, and facilitate their integration into the global economy.
concordia
Concordia is a Python library for constructing and using generative agent-based models to simulate interactions in physical, social, or digital spaces. It facilitates defining environments using a Game Master pattern, where agents act via natural language and the GM translates actions into implementations.
Awesome-GUI-Agent
This repository provides a curated list of papers, projects, and resources focused on multi-modal Graphical User Interface (GUI) agents. It aims to build a comprehensive overview for developing digital assistants capable of interacting with screens.
ralph-claude-code
Ralph enables continuous autonomous AI development cycles using Claude Code, allowing iterative project improvement until completion. It includes intelligent exit detection, rate limiting, and session management to prevent infinite loops and API overuse.
ai-prompts
Instructa AI Prompts is an open-source repository for collecting and sharing AI prompts, best practices, and curated rules for developers. It aims to streamline development workflows by providing ready-to-use prompts for various AI coding environments.
Mysti
Mysti is a VS Code extension that unifies multiple AI coding assistants like Claude, Codex, Gemini, and GitHub Copilot, allowing them to collaborate. It offers features like multi-agent brainstorming, specialized personas, and flexible permission controls to enhance the coding workflow.
WindowsAgentArena
Windows Agent Arena (WAA) is a scalable platform for evaluating multi-modal AI agents on Windows desktops. It offers a reproducible environment for testing agentic workflows and supports large-scale deployment using Azure ML for rapid benchmarking.
blades
Blades is a multimodal AI Agent framework for the Go language, supporting custom models, tools, memory, middleware, etc. It is suitable for multi-turn conversations, chain-of-thought reasoning, and structured output, among other use cases.
Agently-Daily-News-Collector
Agently Daily News Collector is an open-source project using LLMs and the Agently framework to automatically collect news on any topic. Users simply input a topic, and AI agents handle the entire process, generating high-quality news summaries saved as Markdown or PDF files.
daydreams
AI developers often struggle with context switching, state management, and memory persistence, preventing agents from reaching production. Daydreams is the first AI framework with composable contexts, offering true memory, MCP integration, and a TypeScript-first design to build scalable and maintainable AI agents.
Awesome-LLM-in-Social-Science
This repository curates a comprehensive collection of papers focusing on the intersection of Large Language Models and Social Science. It covers evaluation, alignment, application, surveys, and datasets, with a special emphasis on Psychology and intrinsic values.
ChatSim
ChatSim is a system for editable scene simulation in autonomous driving, leveraging large language model agents. It offers two background rendering methods, McNeRF and 3D Gaussian Splatting, balancing realism and speed.
edsl
EDSL simplifies computational social science and market research using AI. It enables designing and running surveys and experiments with numerous AI agents and large language models, providing replicable results with built-in analysis and collaboration tools.
awesome-azure-openai-llm
This repository offers a comprehensive collection of curated resources for Azure OpenAI, Large Language Models (LLMs), and their diverse applications. It features concise summaries, chronological organization, and active tracking to keep users updated on the latest developments in the field.
codeinterpreter-api
This project provides a LangChain implementation of the ChatGPT Code Interpreter, enabling sandboxed Python code execution via CodeBoxes. It allows for advanced data analysis, charting, and file manipulation, with support for local execution and scalable production deployments.
AgentGuide
AgentGuide is a comprehensive learning and job-seeking guide for AI Agent development, offering a systematic, practical, and career-oriented path from beginner to securing a job offer. It covers core AI Agent technologies like LangChain, RAG systems, and Multi-Agent collaboration, alongside detailed interview preparation and project guidance.
agent
Stakpak is a secure, open-source AI agent designed for DevOps, allowing developers to deploy and run infrastructure directly from the terminal. It provides robust security features like secret substitution and network-level guardrails, ensuring AI interactions with production environments are safe and controlled.
ma-gym
ma-gym provides a collection of multi-agent environments built upon the OpenAI Gym interface, facilitating research and development in multi-agent reinforcement learning. It includes various game-like and task-based environments, as well as wrappers to convert single-agent Gym environments into multi-agent forms for debugging.
vespper
Vespper is an AI-powered on-call engineer designed to automate incident response and root cause analysis. It integrates with popular observability and incident management tools to provide real-time contextual insights, helping engineers resolve issues faster.
quark-engine
Quark Engine is a powerful tool for analyzing Android malware, identifying signature behaviors, and generating detailed reports. It helps in understanding various malware families like DroidKungFu, SpyNote, and DawDropper by leveraging a comprehensive rule database.
voice-ai
Rapida is an open-source, end-to-end voice orchestration platform for designing, building, and deploying scalable voice agents. Built in Go with gRPC, it offers real-time audio processing, LLM-agnostic architecture, and full observability for production-grade workloads.
agentsilex
AgentSilex is a transparent, minimal, and hackable agent framework for developers seeking full control over their AI agents. It prioritizes clarity, allowing users to understand and customize every aspect of their agent's operation without hidden complexity.
palico-ai
Palico AI is a comprehensive tech-stack designed to streamline the iterative development of Large Language Model (LLM) applications. It provides tools for building flexible AI agents, previewing changes, improving performance through experiments, and debugging with detailed telemetry.
pydantic-deepagents
Pydantic-deep is a deep agent framework built upon pydantic-ai, extending it with advanced capabilities like planning, file system management, and subagent functionalities. It offers rich toolsets, structured output, and context management for developing sophisticated AI applications.
hive
Hive is a framework for building reliable, self-improving AI agents without hardcoding workflows. It defines goals through conversation with a coding agent, which then generates a node graph with dynamically created connection code.
skyagi
SkyAGI is a Python package that leverages Large Language Models to simulate believable human behaviors. It implements the Generative Agents concept to create an engaging role-playing game experience with highly human-like NPC responses.
LLM-VM
The Anarchy LLM-VM is an optimized backend designed to run open-source LLMs with modern features like tool usage and persistent memory. It acts as a virtual machine for human language, coordinating models, data, prompts, and tools to optimize batch calls and support various architectures.
Odyssey
Odyssey is a novel framework that empowers Large Language Model (LLM)-based agents with open-world skills for exploration in Minecraft. It introduces a comprehensive skill library, a fine-tuned LLaMA-3 model, and a new benchmark for evaluating agent capabilities.
vizra-adk
Vizra ADK is a comprehensive Laravel package for building autonomous AI agents that can reason, use tools, and maintain persistent memory. It allows developers to create intelligent, interactive agents that integrate seamlessly with their Laravel applications.
axar
AXAR AI is a lightweight TypeScript framework for building robust, production-ready LLM-powered agentic applications. It emphasizes explicit control, type-first design, and familiar coding practices to simplify debugging and integration into existing workflows.
agents-aea
The AEA framework allows users to create Autonomous Economic Agents that represent individuals, organizations, or objects in the digital world. These agents operate autonomously to look after their owner's interests and create economic value with minimal interference.
Agent-Framework-Samples
This repository offers a comprehensive guide to building intelligent agents with the Microsoft Agent Framework. It provides practical examples and tutorials for both Python and .NET implementations, covering core concepts, tool integration, RAG, and multi-agent systems.
GPA-LM
This repository serves as a comprehensive collection of research papers focused on the intersection of game-playing agents and large language models. It covers various methods, applications, and challenges, providing a dynamic timeline of recent advancements in the field of LLM-based agents for games.
MAgent
MAgent is a research platform designed for many-agent reinforcement learning, supporting scaling from hundreds to millions of agents. It provides a robust environment for exploring artificial collective intelligence, offering compatibility with Linux and OS X and various deep learning frameworks.
pytorch-DRL
Pytorch-madrl provides modular PyTorch implementations for a range of Deep Reinforcement Learning (DRL) algorithms, suitable for both single and multi-agent systems. It features a unified agent interface with components for environment interaction, training, and action selection to promote code reusability across different DRL methods.