Meilleurs Agents IA CLI
35 agents trouvés
Agents natifs terminal pour développeurs.
Devin
Devin by Cognition est l'agent de codage autonome pionnier. Il peut planifier, écrire, déboguer et déployer des projets logiciels complets de bout en bout avec une intervention humaine minimale, fonctionnant comme un développeur junior.
OpenAI Codex
OpenAI Codex est un agent de codage autonome intégré à ChatGPT Pro. Il s'exécute de manière asynchrone dans un environnement cloud en bac à sable, gérant plusieurs tâches de codage en parallèle pendant que vous vous concentrez sur d'autres travaux.
GitHub Copilot Agent
GitHub Copilot Agent gère de manière autonome les tâches de génie logiciel assignées via les issues GitHub. Il crée des branches, écrit du code, exécute des tests et soumet des PR sans intervention humaine.
Cursor Agent
Cursor Agent Mode transforme l'IDE en un environnement de codage autonome. Il peut planifier des modifications en plusieurs étapes, éditer plusieurs fichiers, exécuter des commandes terminal et itérer jusqu'à ce que la tâche soit terminée.
Bolt.new
Bolt.new by StackBlitz génère des applications web complètes à partir de prompts en langage naturel directement dans le navigateur. Il gère de manière autonome l'échafaudage du projet, le codage, la gestion des dépendances et le déploiement.
SWE-Agent
SWE-Agent by Princeton NLP est un agent autonome open-source conçu pour résoudre des issues GitHub réelles. Il a obtenu les meilleurs résultats sur SWE-Bench et fournit une base pour construire des agents de codage personnalisés.
CrewAI
CrewAI est un cadre d'orchestration multi-agent leader qui vous permet de définir des équipes d'agents IA avec des rôles, des objectifs et des outils spécifiques. Les agents collaborent sur des tâches complexes comme une véritable équipe.
LangGraph
LangGraph by LangChain est un cadre pour construire des applications multi-agents avec état sous forme de graphes. Il prend en charge des flux de travail complexes avec des cycles, des branchements et une intervention humaine.
AutoGen
AutoGen by Microsoft Research permet de construire des systèmes multi-agents où les agents conversent, collaborent et exécutent des tâches ensemble. Il prend en charge la participation humaine et des comportements d'agent personnalisables.
SuperAGI
SuperAGI est une plateforme open-source pour construire et déployer des agents autonomes avec une place de marché pour les outils et les modèles d'agents. Elle fournit une infrastructure pour la gestion du cycle de vie des agents.
OpenAI Agents SDK
OpenAI Agents SDK est le cadre officiel pour construire des agents IA de qualité production. Il fournit des primitives pour l'utilisation d'outils, les transferts entre agents, les garde-fous et le traçage.
Anthropic Computer Use
Anthropic Computer Use permet à Claude de contrôler un ordinateur de manière autonome — cliquer, taper, naviguer dans les applications et accomplir des tâches comme un humain le ferait sur n'importe quelle application de bureau.
Tabnine
Tabnine is an AI code assistant that provides intelligent code completions, explanations, and code generation. It offers unique privacy features with models trained on your codebase, and supports all major IDEs with on-premise deployment options.
Codeium (Windsurf)
Codeium, now part of Windsurf, provides free AI-powered code completions, chat, and search across 70+ languages. Its Cascade feature enables multi-file editing with agentic workflows, rivaling paid alternatives like GitHub Copilot.
Amazon Q Developer
Amazon Q Developer is AWS's AI-powered coding assistant integrated into IDEs and the AWS console. It generates code, transforms legacy applications, performs security scans, and helps architect AWS solutions with deep cloud service knowledge.
Aider
Aider is an open-source AI pair programming tool that works in your terminal. It integrates with your local git repository, makes multi-file edits, runs tests, and commits changes — working directly with your existing codebase and development workflow.
Cline
Cline is an autonomous AI coding agent that runs as a VS Code extension. It can create files, execute terminal commands, use a browser, and interact with external APIs — functioning as a full development assistant within your IDE.
Qodo (CodiumAI)
Qodo (formerly CodiumAI) is an AI coding agent focused on code integrity and quality. It automatically generates comprehensive test suites, analyzes code behavior, detects edge cases, and suggests improvements — ensuring your code is robust before it ships.
AgentGPT
AgentGPT is an open-source platform that lets you create and deploy autonomous AI agents directly in your browser. Define a goal and the agent autonomously breaks it into subtasks, executes them, and delivers results — no coding required.
AutoGPT
AutoGPT is one of the first and most popular open-source autonomous AI agent projects. It chains LLM calls to autonomously achieve any goal you set — breaking tasks into subtasks, using tools, and iterating until completion.
BabyAGI
BabyAGI is a minimalist task-driven AI agent that creates, prioritizes, and executes tasks autonomously. Starting from a single objective, it generates a dynamic task list, prioritizes by importance, and executes each task with context from previous results.
MetaGPT
MetaGPT simulates a complete software company with AI agents playing roles of product manager, architect, engineer, and QA. Given a one-line requirement, the agents collaborate to produce complete software with PRD, design, code, and tests.
ChatDev
ChatDev creates a virtual software development company where AI agents play CEO, CTO, programmer, tester, and designer roles. Through structured chat-based collaboration, they produce complete software projects from a simple description.
Agency Swarm
Agency Swarm is an open-source framework for creating teams of collaborative AI agents. Each agent has specialized tools and roles, communicating through defined channels to accomplish complex tasks that require multiple expertise areas.
Flowise
Flowise is an open-source visual tool for building LLM applications and AI agents. Its drag-and-drop interface lets you create complex agent workflows, RAG pipelines, and chatbots without writing code — using LangChain and LlamaIndex under the hood.
n8n AI
n8n is an open-source workflow automation platform with powerful AI agent capabilities. It combines traditional automation with LLM-powered nodes — enabling you to build AI agents that interact with 400+ services while maintaining full data control.
Haystack
Haystack by deepset is an open-source framework for building production-ready NLP pipelines and search agents. It excels at building RAG systems, question-answering agents, and semantic search applications with modular, composable components.
Marvin AI
Marvin AI is an open-source framework that makes building AI agents and tools simple and reliable. It provides structured outputs, AI-powered functions, and task orchestration primitives that turn natural language into typed, validated Python objects.
DSPy
DSPy by Stanford NLP is a framework for programming — rather than prompting — language models. It provides composable modules and automatic optimizers that tune prompts, few-shot examples, and model weights to maximize pipeline quality.
Guidance AI
Guidance AI by Microsoft Research is a framework for controlling language model generation. It provides a programming model where you interleave generation, prompting, and logical constraints to produce structured, reliable outputs from any LLM.
Windmill
Windmill is an open-source developer platform that transforms scripts into internal tools, workflows, and UIs. It runs Python, TypeScript, Go, and Bash scripts as webhooks, scheduled jobs, or UI-triggered workflows with auto-generated admin interfaces.
AgentQL
AgentQL is an AI-powered SDK for web scraping and browser automation. It uses natural language selectors instead of CSS/XPath to find elements on any webpage, making web automation resilient to website changes and layout variations.
Skyvern
Skyvern is an open-source AI browser agent that automates web workflows using LLMs and computer vision. Unlike traditional scrapers, it visually understands web pages and adapts to changes — automating complex multi-step web interactions programmatically.
Playwright MCP Agent
Playwright MCP Agent uses Microsoft's Playwright framework with MCP (Model Context Protocol) to let AI models directly control web browsers. It enables any LLM to navigate, click, type, and extract data from websites programmatically.
STORM by Stanford
STORM (Synthesis of Topic Outlines through Retrieval and Multi-perspective question asking) is a Stanford NLP research agent that autonomously researches any topic by simulating multi-perspective conversations with search engines, then generates comprehensive Wikipedia-style articles with full citations.