Top 50 free & open-source AI agents
The complete leaderboard of autonomous agents that can run your business — every one free and open-source. Ranked by live GitHub stars, refreshed automatically every day.
Star counts last synced from GitHub on
Love one of these but don't want to set it up? Our team deploys any agent on this list into your business.
Deploy it for me →Personal AI assistant on your own devices. Multi-channel (25+ platforms: WhatsApp, Telegram, Slack, Discord, Signal, iMessage), local-first gateway, voice, Canvas, cron, browser control, 5,700+ community skills, self-writing skills. Runs 24/7.
Fair-code workflow automation with native AI agent nodes. 400+ integrations, visual builder, LangChain integration, self-hostable. Automates business processes with AI-driven decisions. $1B valuation.
By Nous Research. Self-improving AI agent with persistent cross-session memory, skills system, 22 messaging platforms, cloud-native 24/7 operation. Learns from past actions. MIT license.
The pioneer that started the AI agent movement. Goal decomposition, autonomous execution, block-based visual builder, marketplace of pre-packaged agents, production-grade self-hosting. 184k+ stars.
By SST/Anomaly. Provider-agnostic terminal coding agent. 75+ LLM providers, fully offline, 6.5M monthly devs. GitHub Copilot partnership. Go/Typescript. Best open-source coding agent.
Visual AI agent and RAG workflow builder. Drag-and-drop multi-agent flows. Native LangChain, MCP, and A2A protocol support. Low-code agent construction.
Production-ready LLM app platform. Visual agent workflow builder, RAG pipelines, 50+ built-in tools, model management, observability. Agentic workflows with human-in-the-loop.
The foundational agent framework. Modular components for chains, agents, memory, retrieval, tool use. 100M+ PyPI downloads. Powers most agent ecosystems.
Google's open-source terminal coding agent. Gemini 3.1 Pro, 1M token context, free tier. ReAct loop, MCP support, file ops, web fetch. Apache 2.0.
YC W25. Makes websites accessible for AI agents. Real browser automation with Playwright. 1,200+ community automations, trained models for browser tasks. The default browser layer for the agent ecosystem.
(Formerly OpenDevin). $18.8M Series A. Autonomous software engineer. 72% SWE-Bench, solves 50%+ GitHub issues. SDK, CLI, cloud platform. 100+ LLM providers. MIT license.
AI agent company orchestration. Org charts, budgets, governance, heartbeats. If OpenClaw is an employee, Paperclip is the company. Bring your own agents, assign goals, track costs. MIT license.
Autonomous financial data analysis platform. Real-time trading insights, AI-powered analysis. Investment professionals' choice. 65K+ stars.
Multi-agent framework simulating a software company. PM, architect, engineer agents. One-line requirement → user stories, competitive analysis, APIs, code. ICLR 2024 oral.
Natural language control of your local machine. Executes Python, bash, browser commands. Runs code locally. Supports Ollama, LM Studio for fully local operation.
VS Code extension coding agent. Full terminal + browser access. Plan multi-step tasks, edit files, execute commands with user permission. Ollama/LM Studio support.
By Microsoft Research. Multi-agent conversation framework. Pioneer of multi-agent orchestration. Now in maintenance mode → succeeded by Microsoft Agent Framework. 58k stars, 559 contributors.
Open-source general AI agent framework. Modular architecture with planning, tool use, autonomous execution. MIT licensed. Rapid star growth.
Open-source drag-and-drop LLM agent builder on LangChain. Visual node graph, deploy as API, embeddable chat widget. Supports all major LLMs.
Role-based multi-agent framework. Define agents with roles/goals, assemble crews. 2B agent executions/year, 27M PyPI downloads. 60%+ Fortune 500. $44M funding.
By Block. Open-source, extensible AI agent that automates engineering tasks end-to-end — writes code, runs commands, and works with any LLM via MCP. Built in Rust.
AI pair programming in terminal. Git-aware, any LLM. Autocompletion, debugging, task automation. One of the most established open-source coding agents.
Ultra-lightweight personal AI agent. Persistent workflows, multi-channel (WebUI, Telegram, Slack, Discord), memory, MCP tools, model routing. Python-based. MIT license.
Graph-based agent orchestration framework. Stateful directed graphs, checkpointing, streaming, human-in-the-loop. Official LangChain agent runtime for complex workflows.
Event-driven agentic orchestration. Durable, resilient execution for apps and AI agents. Battle-tested at Netflix, Tesla, LinkedIn, J.P. Morgan. 30K+ stars.
Tool integration layer for AI agents. 1000+ toolkits, auth management, sandboxed workbench. Connects agents to real APIs (Jira, HubSpot, GitHub, etc).
Hugging Face's code-first, zero-bloat agent framework. Agents write Python, Python runs. Minimal abstraction. First-class Hub integration. Best for local models.
#1 open deep research agent (CMU DeepResearchGym). Autonomous web/local research, detailed reports with citations. MCP server. Any LLM. Apache 2.0.
Fully autonomous web browsing agent. "Fully Local Manus AI." Data extraction, form filling, web-based task automation. Runs entirely locally.
Open-source Zapier alternative with AI agents. 280+ MCP servers, self-hosted, visual workflow builder. AI Support Agent, Smart Email Agent, Daily Report Agent.
AI browser agent with computer vision. Automates complex web workflows using LLMs + vision. Handles text and visual elements. Good for enterprise RPA replacement.
Princeton NLP. Resolves real GitHub issues autonomously. Pioneered Agent-Computer Interface. Clean, minimal, rigorous benchmarks. MCP support. MIT.
Autonomous multi-agent framework. Discord/Telegram/Farcaster connectors, RAG support, modern web dashboard. TypeScript-based. Community-driven.
Agent Operating System in Rust. 137K LOC, 14 crates. Single ~32MB binary. 7 autonomous "Hands" (Lead, Intel, Researcher, Twitter, Browser, etc). 40 channel adapters, 53+ tools, WASM sandbox. Built-in scheduling.
Dev-first autonomous AI agent framework. Build, manage & run autonomous agents. Action console, multiple vector DBs, multi-modal agents, agent trajectory fine-tuning.
First multi-agent framework for scalable agent systems. Role-based simulation, collaborative reasoning. Apache 2.0. Extensive tooling for agent communication.
Self-evolving agent: ~3K lines core code. 9 atomic tools, skill tree that grows with use. 6x less token consumption. Desktop GUI. Controls browser, terminal, filesystem, keyboard/mouse.
Microsoft's production-grade agent framework. Python + .NET. Graph-based workflows, multi-agent, checkpointing, human-in-the-loop, OpenTelemetry. Enterprise-ready successor to AutoGen + Semantic Kernel.
Multi-agent harness for production. Self-healing graph execution, crash recovery, state persistence, cost enforcement, human-in-the-loop. Apache 2.0. 210+ contributors.
Zero-code LLM agent framework. Natural language-driven agent building. Self-developing, self-managing workflow generation. Matches Deep Research with Claude 3.5.
24/7 AI employee team. Low-code multi-agent with handoffs, guardrails, memory, RAG, 100+ LLM providers. Automates complex challenges.
Open-source framework for building AI assistants with custom tools. Modular architecture for production-ready AI agents. Both open-source and managed service available.
TypeScript-native multi-agent. Goal → task DAG automatically. 12 built-in providers, 6 built-in tools + MCP, streaming. Three runtime dependencies. MIT.
Zero-code platform for production-grade AI agents. A2A collaboration, layered memory, progressive skill disclosure, MCP ecosystem, agent marketplace, multi-tenancy. MIT.
Community fork of AutoGen continuing active development. Multi-agent conversations, async, event-driven. Plug-in replacement for AutoGen with new features.
Autonomous AI agent with evolving self-model (ego, affect, identity). Business launcher, agent swarm, crypto payments, prediction markets, 200+ tools. Runs freelancing business autonomously.
AI agent orchestration OS. CEO orchestrator, persistent memory (MemPalace), Critic Loop, atomic budgets, Task DAG, War Room live agent grid. AGPL-3.0 (Community Edition).
Enterprise AI agent orchestration. 24 specialist agents across Finance, HR, Marketing, Operations, Back Office. 1000+ integrations (Composio), real API endpoints, voice agents, RAG.
AI operating system for business. AI CEO + 7 pillar agents (Engineering, Product, Data, Revenue, Customer Success, Operations, People). Bead-chain immutable audit trail. Self-hosted.
Open-source orchestration for zero-human companies. Org charts, budgets, governance, heartbeat execution. MIT license. Bootstrap: `npx ai-ceo onboard`.
📊 Ranking methodology
Agents are ranked primarily by GitHub stars (public community signal), then weighted by autonomous business capability, production-readiness, community activity, and business-automation features. Sources: GitHub, community directories, and verified repositories (June 2026). Selection criteria: must be free & open-source (MIT, Apache-2.0, AGPL, or fair-code), have genuine autonomous agent capabilities, and be applicable to running business operations. Star counts may have changed since publication.
Want one of these agents working in your business?
These are powerful — but self-hosting, integrating, and securing them takes real engineering. We install, connect, and train any agent on this leaderboard for your team, then hand it over running.