Open-source Python framework for orchestrating role-playing, autonomous multi-agent systems. Crews enable collaborative agent teams with defined roles, goals, tools, and tasks; Flows provide event-driven workflow control with state, branching, and routing. Standalone (no LangChain dependency). Supports MCP, A2A, memory, checkpointing, and async execution.
Open-source SDK from Microsoft that unifies AutoGen's multi-agent orchestration with Semantic Kernel's enterprise readiness. Supports Python, C#/.NET, and Go. Provides graph-based workflows, middleware, hosted agents on Foundry, OpenTelemetry observability, and declarative YAML agents.
Agent engineering platform providing modular primitives for building LLM-powered applications. Core value is breadth: swap model providers with one line, compose chains and agents from modular components, prototype to production without switching frameworks. Includes Deep Agents for planning and subagent delegation.
Graph-based framework for building stateful, multi-actor LLM applications on top of LangChain. Models agent execution as a graph of event handlers where each step emits and receives typed events, enabling nested and parallel agent pipelines. Embeds into Python scripts, notebooks, and REST APIs via Starlette/FastAPI.
Open-source context API for searching, scraping, and interacting with the web at scale. Covers 96% of the web including JS-heavy pages. Returns clean markdown, HTML, screenshots, or structured JSON. Includes search, scrape, crawl, map, batch scrape, and an AI agent for autonomous data gathering. 150K+ GitHub stars.
Browser infrastructure platform that makes the web programmable for AI agents. Provides managed Chromium sessions with persistent cookies/localStorage, session recordings for debugging, and Playwright/Puppeteer-compatible APIs. Drops in as a browser provider for agent-browser and other tools.
Open-source Python library that makes websites accessible for AI agents. Provides a high-level Agent abstraction that controls a browser to complete tasks like searching, clicking, form-filling, and extraction. Works with OpenAI, Anthropic, and Browser Use's own optimized model. Converts web UI into structured browser states agents can reason over.
Fast native Rust CLI for browser automation designed for AI agents. Issues simple CLI commands (click, fill, navigate, screenshot) that agents can call programmatically. Supports session isolation, state restore/save, CDP 연결, proxy, extensions, and many browser configuration options. Integrates with Browserbase as an engine provider.
Flexible Python framework for building AI agents compatible with multiple LLM providers. Provider-agnostic implementation enables maximum portability across OpenAI, Anthropic, and other backends. Part of Open Agent Tools, a collection of open-source utilities for AI agent development, trading systems, and developer tooling.
Batteries-included agent framework from LangChain built on top of LangGraph. Deep Agents can plan complex tasks, spawn subagents for delegation, and use a file system for context management. Designed for multi-step reasoning tasks like research, code analysis, and document synthesis where simple single-pass agents fall short.