Agents

Autonomous AI systems that perceive, reason, and act in the world. These agents represent the closest things we have to AGI-like behavior in narrow domains -- writing code, conducting research, operating computers, and navigating virtual worlds.

Coding & Software Engineering Agents

Name	Github Stars	Introduction	Notes
Aider		AI pair programming in your terminal. Edit code across your entire repo with LLMs. Tops SWE-bench Lite.	Best-in-class AI pair programmer. Works with GPT-4o, Claude, DeepSeek, local models.
SWE-agent		Turns LLMs into software engineering agents that fix real GitHub issues. Pioneered the Agent-Computer Interface (ACI) concept.	Paper. Princeton NLP. Key SWE-bench benchmark agent.
goose		Open-source extensible AI agent that goes beyond code suggestions — install, execute, edit, and test with any LLM. Written in Rust.	Supports MCP and ACP protocols. By Block.
Open Interpreter		A natural language interface for computers. Lets LLMs run code (Python, JS, shell) locally with no restrictions.	Full computer control via natural language. Like ChatGPT Code Interpreter but unrestricted.
Cursor Composer 2		Next-generation IDE incorporating Kimi K2.5 model. Beats Anthropic Opus 4.6 on major coding benchmarks at lower cost.	AI-native IDE with frontier model integration.
Opencode		Open source AI coding agent using most models including free and local. Terminal, desktop, IDE, or extension. Privacy-sensitive environments.	Multi-model, privacy-first coding agent.
Claude Review		Code review on every pull request that Claude Code makes. Research preview for Claude Teams and Enterprise.	Automated PR review for AI-generated code.

Research & Knowledge Agents

Name	Github Stars	Introduction	Notes
STORM		Stanford: Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking. Writes Wikipedia-quality long-form articles autonomously.	Paper. Expert-level knowledge synthesis agent.
AI Scientist		Fully autonomous scientific research pipeline: generates ideas, implements experiments, writes and reviews full academic papers.	Paper. By Sakana AI. ASI-threshold capability.
gpt-researcher		GPT-based autonomous agent for comprehensive online research on any topic. Searches, reads, and synthesizes long-form reports.	-
mem0		Universal memory layer for AI agents. Persistent long-term, episodic, and semantic memory across agent sessions.	52k+ stars. Key missing component between current agents and AGI.
Autoresearch		Automates the scientific method with AI agents. Runs hundreds of ML experiments per night - experiment, results, modify code, repeat.	By Andrej Karpathy. ASI-threshold automation.
Claude Code Channels		Experimental feature allowing communication with Claude using Telegram or Discord. Competes with OpenClaw.	Multi-platform Claude integration.
Claude Cowork Dispatch		Control Claude Cowork from your phone. Claude runs on computer, assign tasks from anywhere, get text notification when done.	Remote Claude control via mobile.

Computer-Use & Desktop Agents

Name	Github Stars	Introduction	Notes
everything-claude-code		Agent harness performance optimization system with skills, instincts, memory, and security for AI coding agents.	Cognitive scaffolding for agentic AI — directly relevant to persistent AGI.
nanobot		Ultra-lightweight personal AI agent. Minimal footprint with full agent capabilities.	39k+ stars. Edge-AGI and ubiquitous deployment.
Screenpipe		Run agents that work for you based on what you do. Continuously observes your screen, builds personal memory, and triggers AI actions.	Always-on environmental awareness — a prototype AGI assistant.

Local Agent Stacks

Local-first agent orchestration stacks that replace cloud-based model dependence with on-device deployment.

Name	Introduction	Notes
Qwen-3-coder	Local agent stack for coding that can replace cloud-based orchestration tools.	Local-first coding agent infrastructure.
Ollama	Run Llama, Mistral, and other models locally. Simple CLI for local model deployment and inference.	Local model serving for agents.
goose	Already listed in coding agents - local agent framework supporting MCP and ACP protocols.	Local agent framework (see Coding & Software Engineering Agents).

Embodied & Simulation Agents

Name	Github Stars	Introduction	Notes
generative_agents		Generative Agents: Interactive Simulacra of Human Behavior. 25 AI agents in a simulated town with unique identities, memories, and behaviors.	Paper. Stanford & Google. Emergent social cognition.
Voyager		Open-Ended Embodied Agent with LLMs. Plays Minecraft autonomously, continuously discovers new skills via a self-growing code library.	Paper. NVIDIA/CMU. Lifelong skill acquisition.
RoboGen		Generative simulation for automated robot learning. Auto-generates tasks, environments, and training data using GPT-4.	Paper. Self-improving robot curriculum.
ai-town		Deployable starter kit for building AI town — a virtual town where AI characters live, chat, and socialize. By a16z.	-

Enterprise & Agentic AI Platforms

AI agents moving from demos to enterprise production. These platforms enable organizations to deploy autonomous agents into real workflows at scale.

Name	Introduction	Notes
Salesforce Agentforce	Enterprise AI agent platform integrated into Salesforce CRM. Autonomous agents handle sales, service, marketing, and commerce workflows.	Largest enterprise AI agent deployment. Announced 2024.
Microsoft Copilot Studio	Low-code platform for building and deploying custom AI agents and copilots across Microsoft 365 and Azure.	Powers enterprise-wide agentic automation.
Zapier AI Agents	No-code AI agents that automate cross-app workflows. Connects 7,000+ apps with autonomous decision-making and action.	Enterprise workflow automation at scale.