Agents

macOS computer-use agent that reads the screen with OCR, uses a small classifier to choose each click and a writer only for text.

Open-source agent (MIT) that works with any OpenAI-compatible model, keeps a 16-tool core and loads skills on demand

Research agent from Princeton and Stanford that lets a language model fix GitHub issues and other tasks by using tools on its own

Autonomous research agent that reads papers, runs experiments and compiles a LaTeX report with no human in the loop

Lightweight terminal coding assistant with a tool loop and TUI, offered as TypeScript, Python and Rust versions

Open-source desktop computer-use agent that operates apps from natural-language requests.

Autonomous penetration testing agent with a Kali Docker sandbox, a Textual TUI and any OpenAI-compatible model gateway

Python terminal coding agent with DeepSeek by default, MCP tools, sub-agents, approval prompts and reusable learned skills

Terminal coding agent with permission rules, OS sandboxing, MCP, skills and sub-agents, for Anthropic, OpenAI-compatible, Gemini and local models

Single Rust binary for an autonomous, self-improving agent with TUI, CLI and daemon modes and connections to many chat channels

Autonomous penetration testing agent that drives Claude Code or Codex through staged recon, exploit and report pipelines.

Coding agent tuned for small local models, built on the pi agent harness and run from the terminal.

Terminal coding agent built on DeepSeek Harness, with a background host that keeps sessions running and hands tasks between sessions

AI penetration testing agent for authorized assessments, with scope enforcement and a cloud module covering major providers

Terminal coding agent for DeepSeek with MCP, skills and workflows; the README says the repository is no longer maintained

Python rebuild of Claude Code with multi-agent teams, output-compression filters, a nano mode and a web UI

Terminal pentesting agent that runs specialist subagents over a shared engagement state and supports 20+ LLM providers

Open-source reverse-engineering agent that uses Ghidra and LLMs to rebuild and validate C/C++ functions from binaries.

Visual workbench for AI coding agents that shows every model call, with sessions run locally or on a server

Open-source terminal coding agent for DeepSeek V4 that keeps prompt prefixes byte-stable for cache-friendly, low-cost sessions

Runs coding agents on your own computer and lets you message them as chat bots from a phone, desktop or terminal.

Open-source agent harness from MindsHub that runs as the default agent in its Cowork app and standalone in a terminal.

Windows desktop AI companion built on Electron with a Live2D character, long-term memory, voice and tool calling.

tmux-based multiplexer that runs coding agents in separate git worktrees and merges or opens a PR when a task is done.