Top AI agents
Find AI agents, frameworks and tools for coding, research, automation and more.
The Top 50 highlights the most-starred indexed projects. Rising alternatives surfaces projects outside that list gaining stars in the measured period.
Explore the top 50 Explore rising alternatives Star this directory
Explore by use case
- CodingAgents that assist software development.
- ResearchSearch, investigate and synthesize evidence.
- Browser automationAgents and infrastructure for browser and web interaction.
- Knowledge and memoryRetrieval, knowledge and persistent agent memory.
- Workflow and business automationCompose and operate business workflows.
- FrameworksLibraries for building and coordinating agents.
- Personal agents and workspacesPersonal assistants and conversational workspaces.
- SkillsReusable instructions and skill packages.
- Domain-specificSpecialized research, education and industry agents.
- Agent operationsOperate, connect and manage agents.
Top 50 by GitHub stars
Initial edition: 50 independently reviewed projects, ordered by GitHub stars; not a global ranking.
| Project | Utility | Kind | Stars | Starred in September | Introduction |
|---|---|---|---|---|---|
| Superpowers | Skills | skill | 309k+ | 14k+ | A composable skills framework for coding agents centered on clarification, design, planning, testing, and review. It changes an existing agent's workflow rather than supplying a new model. |
| Hermes Agent | Personal agents and workspaces | agent | 259k+ | 11k+ | A persistent personal agent with terminal, desktop, messaging, and scheduled-work interfaces. It can create reusable skills and search past sessions, with configurable model providers. |
| OpenCode | Coding | agent | 223k+ | 8k+ | A coding agent with terminal and desktop interfaces. Its separate plan and build modes let you explore a repository before giving the agent permission to change it. |
| n8n | Workflow and business automation | platform | 209k+ | 3k+ | A fair-code automation platform combining a visual canvas, custom code, and agent nodes. It can run self-hosted or as a managed cloud service. |
| Firecrawl | Browser automation | infrastructure | 197k+ | 12k+ | A web-data API for search, scraping, document parsing, and interaction. It supplies agent-ready content and structured extraction, with both public code and a hosted service. |
| Anthropic Skills | Skills | skill | 185k+ | 6k+ | A set of reusable task instructions, scripts, and resources for Claude's skills system. These packages extend an existing agent; they are not standalone autonomous agents. |
| Agency Agents | Skills | skill | 165k+ | 6k+ | A collection of specialized agent personas and workflow instructions. It installs into compatible coding agents and provides role-oriented guidance, not verified professional expertise. |
| Dify | Workflow and business automation | platform | 160k+ | 3k+ | A visual platform combining agent workflows, retrieval pipelines, model management, and deployment. Its license adds conditions beyond standard Apache-2.0. |
| Langflow | Workflow and business automation | platform | 156k+ | 1k+ | A visual builder for agents and workflows with API and MCP deployment surfaces. Components can be customized in Python while flows remain inspectable in a graphical interface. |
| Claude Code | Coding | agent | 154k+ | 5k+ | Anthropic's coding tool works from a terminal, IDE, or GitHub integration to explain code and carry out development tasks. Its public repository is not an open-source license for the product. |
| LangChain | Frameworks | framework | 148k+ | 2k+ | A framework for building agents and model-powered applications from interoperable components. It is a development library, not a ready-to-use autonomous assistant. |
| Codex CLI | Coding | agent | 132k+ | 7k+ | OpenAI's local coding agent provides a terminal interface for repository work. The CLI is distinct from Codex's hosted web and desktop experiences. |
| Browser Use | Browser automation | agent | 119k+ | 5k+ | A tool-using browser agent available as a local library/CLI and a hosted cloud service. The agent decides browser actions from task instructions and a configured model. |
| TradingAgents | Domain-specific | framework | 118k+ | 7k+ | A research framework that coordinates analyst, researcher, and trader roles over market information. Its generated decisions are research artifacts, not investment advice. |
| Orca | Agent operations | platform | 111k+ | 24k+ | A workspace for running coding agents side by side in isolated Git worktrees. It combines agent sessions, terminals, review tools, and optional remote/mobile control. |
| Paperclip | Agent operations | platform | 109k+ | 15k+ | An orchestration application for teams of agents, with tasks, goals, budgets, permissions, and history. It manages existing agents rather than replacing their runtimes. |
| Gemini CLI | Coding | agent | 108k+ | 816 | Google's terminal agent is now documented for enterprise Code Assist access. The current README directs free individual and Google AI Pro/Ultra users to Antigravity CLI instead. |
| Claude-Mem | Agent operations | infrastructure | 107k+ | 2k+ | Persistent context infrastructure that captures agent activity and makes summaries available in later sessions. It includes adapters for multiple agent hosts, with local and hosted memory paths. |
| autoresearch | Research | agent | 98k+ | 2k+ | A small setup for agents to iterate on single-GPU language-model training experiments. It is a specialized research loop, not a general web-research assistant. |
| RAGFlow | Knowledge and memory | platform | 93k+ | 1k+ | A retrieval and agent platform for turning documents into usable model context. Its current documentation combines document processing, knowledge compilation, and agent workflows. |
| Crawl4AI | Browser automation | infrastructure | 86k+ | 4k+ | A crawler and scraper that turns websites into Markdown for retrieval systems and agents. It can run in your own Python process, a Docker server, or a hosted service. |
| DeerFlow | Research | agent | 84k+ | 2k+ | DeerFlow 2 is a long-horizon agent harness using subagents, memory, skills, and sandboxes. It is broader than the original research framework, which remains on a separate 1.x branch. |
| LobeHub | Personal agents and workspaces | platform | 83k+ | 1k+ | A workspace for creating, scheduling, and collaborating with agent teammates. It combines conversational interfaces with operator and integration features. |
| MetaGPT | Frameworks | framework | 71k+ | 672 | A multi-agent framework organized around software-team roles and procedures. It turns a requirement into coordinated planning and development artifacts. |
| Cline | Coding | agent | 71k+ | 2k+ | A coding agent available through editor extensions, a CLI, a desktop app, and an SDK. Plan/act workflows and human approvals are central to its documented repository experience. |
| Open Interpreter | Coding | agent | 68k+ | 387 | The current Open Interpreter is a coding agent focused on harness configurations for lower-cost models. Its README describes a Codex-derived implementation with selectable harnesses and ACP compatibility. |
| Mem0 | Knowledge and memory | infrastructure | 68k+ | 2k+ | A memory layer for retaining and retrieving information across agent interactions. It offers a library, a self-hosted server, and a managed platform with distinct feature scopes. |
| AnythingLLM | Knowledge and memory | platform | 67k+ | 1k+ | A local-first application for document chat, agent tools, and configurable model connections. Desktop and self-hosted deployments support different operating needs. |
| AutoGen | Frameworks | framework | 61k+ | 597 | A historical multi-agent application framework. Its current README says the project is archived and no longer maintained, even though the live GitHub archive flag was unset at collection. |
| CrewAI | Frameworks | framework | 60k+ | 1k+ | A Python framework combining role-based agent crews with event-driven flows. It provides abstractions for collaborative agents while leaving tools and application logic to developers. |
| Marketing Skills | Skills | skill | 57k+ | 5k+ | Reusable skills for conversion work, copywriting, SEO, analytics, and growth engineering. They target compatible agents and include optional guided connections to marketing tools. |
| Hindsight | Knowledge and memory | infrastructure | 56k+ | 21k+ | A memory system exposing retain, recall, and reflect operations for agents. It organizes memory into banks and supports self-hosted, embedded, and cloud deployment options. |
| goose | Coding | agent | 56k+ | 1k+ | A general-purpose agent with desktop, CLI, and API interfaces. It combines model-provider choice with extensible tools, including MCP integrations, for coding and other local workflows. |
| Scientific Agent Skills | Domain-specific | skill | 55k+ | 6k+ | A library of scientific workflows for agents supporting the Agent Skills standard. It covers analysis and database tasks but does not establish that an agent is a qualified scientist. |
| DeepTutor | Domain-specific | platform | 52k+ | 2k+ | A learning workspace connecting tutoring, problem solving, quizzes, research, and reading. Its agents can reuse approved learning context across multiple modes. |
| nanobot | Personal agents and workspaces | agent | 49k+ | 1k+ | A self-hosted personal agent with browser, terminal, and messaging interfaces. It supports file work, remembered preferences, and scheduled tasks around a configured model. |
| CowAgent | Personal agents and workspaces | agent | 47k+ | 540 | A personal assistant harness combining planning, skills, memory, knowledge, and multi-agent features. It can run on a computer or server and connect to messaging channels. |
| LibreChat | Personal agents and workspaces | platform | 46k+ | 2k+ | A self-hosted chat and agent workspace supporting multiple model providers, tools, and skills. Its current release also documents managed agent APIs and experimental attached code workspaces. |
| agent-browser | Browser automation | infrastructure | 44k+ | 1k+ | A native browser-automation CLI designed to be called by AI agents. It supplies browser commands and snapshots rather than a complete autonomous agent on its own. |
| LangGraph | Frameworks | framework | 44k+ | 1k+ | A low-level orchestration framework for long-running, stateful agent workflows. It provides graph control, persistence, and human-in-the-loop mechanisms rather than a finished assistant. |
| Agno | Frameworks | framework | 42k+ | 508 | An agent platform and SDK ecosystem for building agents, teams, and workflows. Its AgentOS layer exposes agents as services with operational controls. |
| BettaFish | Research | agent | 42k+ | 287 | A multi-agent system for researching public opinion and social-media material. Its documented workflows combine collection, analysis, and report generation rather than just a chat interface. |
| Vane | Research | platform | 37k+ | 471 | A self-hostable answering engine combining web search, uploaded documents, and local or cloud models. It exposes source-cited answers through a browser interface. |
| Composio | Agent operations | infrastructure | 30k+ | 470 | An SDK and service interface for connecting agents to authenticated application tools. It provides per-user sessions, integrations, triggers, and provider adapters. |
| Page Agent | Browser automation | framework | 29k+ | 437 | A JavaScript agent that operates inside a web page using text-based DOM interaction. Its base integration does not require a headless browser or screenshots. |
| Stagehand | Browser automation | framework | 25k+ | 1k+ | A browser-agent SDK for natural-language actions, observation, and structured extraction. It lets application code combine ordinary browser control with model-assisted steps. |
| Activepieces | Workflow and business automation | platform | 25k+ | 691 | An automation platform with a visual builder and extensible TypeScript integrations called pieces. Its integrations can also be exposed as MCP tools for agents. |
| Agent Development Kit | Frameworks | framework | 22k+ | 368 | Google's code-first Python toolkit for building, evaluating, and deploying agent workflows. It is optimized for Gemini but documents model- and deployment-agnostic operation. |
| Windmill | Workflow and business automation | platform | 18k+ | 348 | A code-first platform that turns scripts into workflows, APIs, scheduled jobs, and applications. It can orchestrate agent tasks while keeping ordinary code and review workflows in the loop. |
| Letta Code | Knowledge and memory | agent | 3k+ | 319 | A stateful agent harness whose context, memory, and skills can evolve across sessions. It supports local CLI interaction and optional remote or messaging access. |
No entries in this selection.
Rising alternatives
| Project | Utility | Kind | Stars | Starred in September | Introduction |
|---|---|---|---|---|---|
| OpenMonoAgent | Coding | agent | 1k+ | 67 | A .NET coding agent with bundled llama.cpp inference and local execution options. It is an alternative for developers who want to run the model on their own hardware rather than defaulting to a cloud coding service. |
| Magic Context | Knowledge and memory | infrastructure | 2k+ | 287 | Memory infrastructure for coding agents that captures project knowledge, consolidates it, and recalls it across sessions. It focuses on background context management around an existing coding agent rather than a separate chat workspace. |
| Mnemon | Knowledge and memory | infrastructure | 676 | 67 | A standalone persistent-memory binary with graph storage and recall tools. It lets the host agent's model supervise memory decisions instead of embedding a separate model in the memory pipeline. |
| tRPC Agent Go | Frameworks | framework | 1k+ | 82 | A Go framework for building tool-using agents and multi-agent workflows. It provides a Go-native alternative to the Python-oriented frameworks in the featured list. |
| Aegra | Agent operations | infrastructure | 1k+ | 73 | Self-hostable infrastructure for serving LangGraph agents through compatible deployment APIs with PostgreSQL persistence. It is a deployment-focused alternative rather than a general agent-management workspace. |
| Open Agent Email | Agent operations | infrastructure | 2k+ | 68 | Self-hosted email infrastructure with per-agent mailboxes, API/SDK access, and approval controls for risky outgoing messages. It focuses on an agent's email identity and conversation flow rather than general application orchestration. |
No qualifying rising alternatives in the measured period.