Ecosystem

Powerful AI coding agent for the terminal. Multi-model support, clean TUI. Pioneered many patterns others adopted.

All-in-one agentic engineering platform. #1 coding agent on OpenRouter with 1.5M+ users and 25T+ tokens processed.

Google's open-source AI agent for the terminal. Brings Gemini directly to your command line with MCP support and 1M token context.

OpenAI's lightweight coding agent for the terminal. Written in Rust for speed.

AI-powered IDE by Codeium. Deep codebase understanding with Cascade, their agentic flow engine.

The AI code editor that started the agentic IDE category. Composer mode, multi-file editing, strong model support.

Official OpenAI framework for multi-agent workflows. Lightweight, built-in MCP tool integration, persistent sessions, and realtime voice agents.

Wraps the Claude Code CLI as a programmable SDK. Bidirectional interactive conversations, custom tools and hooks as functions. Everything Claude Code can do, available programmatically.

Google's official Agent Development Kit. Code-first, multi-language (Python/Go/TS/Java). Optimized for Gemini but model-agnostic. Built-in evaluation and deployment tools.

Provider-agnostic TypeScript toolkit from the Next.js creators. 20M+ monthly downloads. Works with React, Svelte, Vue, Angular. The standard for AI-powered web apps.

Pure Python agents with no graphs, chains, or heavy abstractions. Agents, teams, and workflows as plain Python. High-performance runtime. Formerly PhiData.

Hugging Face's code-first agents in ~1,000 lines of core logic. Agents write actions as code, not just tool calls, reducing steps and LLM calls by ~30%.

Type-safe agent framework built on Pydantic. Leverages Python's type system to catch agent logic errors at dev time. The Pythonic way to build AI agents.

Program language models instead of prompting them. Declarative framework from Stanford NLP. Compilers automatically optimize prompts and weights. Fundamentally different paradigm.

Simulates a software company. Agents take roles (PM, architect, engineer) and follow SOPs. Give it one requirement, get PRD, design, tasks, and code.

The original autonomous agent (2023). Most-starred AI agent repo. Now a full platform for creating, deploying, and managing continuous AI agents.

Stateful agents with OS-like memory management. Agent controls its own memory through function calls (read, write, search, archive). Formerly MemGPT.

Production-ready AI orchestration by deepset. Modular pipelines for RAG, agents, and semantic search. Standardized tool calling with transparent, explicit architecture.

Microsoft's enterprise-grade AI orchestration SDK. Multi-language (C#/Python/Java). Core of the Microsoft Agent Framework. Strong Azure and Copilot integration.

Research-focused multi-agent framework. 'Finding the Scaling Law of Agents.' 100+ researchers contributing. Related projects include OASIS (1M-agent simulations).

Lightweight, modular 'atomic' approach. Single-purpose, composable components built on Instructor + Pydantic. Maximum control with minimal abstraction.

Agent2Agent: open protocol for communication between opaque agentic applications. Linux Foundation project backed by Google.

Google's AI design tool. Connects to your UI designs and generates production-ready code.

The collaborative design platform. AI features, dev mode, and MCP server support make it the bridge between design and code.

Office Compound Engineering plugin for Claude Code, Codex, and more. Structured workflows for engineering teams.

Full skill suite for Claude Code: QA, shipping, browsing, design review, investigation, deployment, and more. The backbone of our workflow.