Agent Harnesses

Factory AI

Enterprise AI coding platform built around autonomous "Droids" that plan, code, test, and open pull requests across a codebase, with human review built into the loop.

SWE-agent

Open-source research agent from Princeton and Stanford that gives an LLM a custom agent-computer interface to fix real GitHub issues, scoring strongly on SWE-bench.

Devika

Open-source agentic engineer that plans a task, researches it with web search, and writes code step by step, explicitly built as an open alternative to Devin.

Pieces for Developers

On-device AI developer workspace for saving, searching, and reusing code snippets, with long-term memory of your work across apps and local model options.

Mastra

Open-source TypeScript agent framework from the team behind Gatsby, with built-in workflows, RAG, memory, and evals for building production AI agents in a JS-native stack.

Aider Chat

Open-source, git-aware AI pair programmer that runs in your terminal, mapping your codebase and auto-committing changes with sensible messages.

talk-codebase

CLI tool for chatting with your codebase using OpenAI, LlamaCpp, or GPT4All, with an offline mode. Archived by its author in February 2025.

GPT Engineer

Open-source CLI that generated full software projects from a single prompt. Archived April 2026; its idea evolved into Lovable, co-founded by its creator.

Kilo Code

Open-source agentic coding platform with five agent modes, support for 500+ models at no markup, an MCP marketplace, and parallel subagents.

Roo Code

VS Code extension forked from Cline, adding customizable agent modes including one for sysadmins. Shut down on May 15, 2026, repo now archived.

OpenCode

Open-source terminal coding agent with a TUI that works with any model provider, including local models, and supports LSP, MCP and custom agents.

Amp

Sourcegraph's agentic coding tool for the terminal and VS Code, with sub-agents, thread sharing and pay-as-you-go usage.

Qwen Code

Open-source command-line coding agent from the Qwen team, adapted from Gemini CLI and tuned for the Qwen3-Coder models.

Crush

Charm's terminal AI coding agent with a polished TUI, multi-provider model support, LSP integration and MCP extensibility.

Claude Agent SDK

Anthropic's SDK for Python and TypeScript that exposes the Claude Code agent loop, tools, hooks and MCP support for building custom agents.

Grok Build

xAI's terminal coding agent with plan mode, parallel subagents with git worktree isolation and headless CI use, powered by the grok-build model.

Refact.ai

Open-source AI coding agent that plans and executes engineering tasks end to end, using RAG for repository-aware completions and self-hosting.

Cursor

AI-first code editor forked from VS Code, with Tab autocomplete, multi-file Agent mode, Composer, and background agents running on Claude, GPT, or Gemini.

Claude Code

Anthropic's agentic coding tool, running in the terminal, IDE, desktop app, and browser, reading your codebase and editing files with explicit approval.

Cosine

AI software engineer that runs from the CLI, cloud, or desktop, retrieving codebase context and planning, writing, and testing changes via its Genie agent.

Plandex

Open-source, terminal-based AI coding agent for large, multi-step tasks, handling up to 2M tokens of context in a review sandbox before applying changes.

ToolHive

Enterprise platform from Stacklok for running MCP servers, isolating each in its own container with minimal permissions and per-request access policy.

OpenAI Codex

OpenAI's cloud-based autonomous coding agent, powered by the GPT-5 family, reading a codebase and opening pull requests from written instructions.

Amazon Q Developer CLI

AWS's terminal agent combining IDE-style autocomplete for CLIs like git and docker with natural-language chat and agentic command execution.

poorcoder

Small set of Bash scripts that bridge web-based LLM chat subscriptions, like Claude.ai or Grok, with your terminal: extract context, apply responses, and write commit messages.

cmd-ai

Natural-language shell command generator with built-in safety filters and explanations, using local Ollama models by default or OpenAI, Gemini, or Claude.

Shell Whiz

AI assistant for the command line that suggests shell commands from a plain-English description and explains what each one does.

Butterfish

Shell wrapper for bash/zsh that adds AI prompting on top of your existing shell. A capital letter asks the LLM a question; a "!" prefix lets it act as an agent.

TmuxAI

Open-source terminal assistant living inside your tmux session, reading all your panes to help without needing a special shell or wrapper.

Warp

Rust-based, GPU-accelerated terminal with Agent Mode that turns plain-language requests into shell commands, plus Oz, a cloud orchestrator for background agents.

Codename Goose

Open-source AI agent from Block, runnable as a desktop app or CLI, connecting to external systems via MCP and now governed by the Agentic AI Foundation.

Smol Developer

Library for embedding a developer agent into your own app, letting it scaffold a full codebase from a single prompt. Around 12,200 GitHub stars.

Aider

Open-source, git-aware AI pair programmer that runs in your terminal. Same project as "Aider Chat" elsewhere in this directory, now maintained at Aider-AI/aider.

Rovo Dev

Atlassian's terminal coding agent, built into the CLI with 300+ commands for querying Jira, Bitbucket, and GitHub data through the Teamwork Graph.

Gemini CLI

Google's open-source terminal coding agent built on Gemini models with a 1M-token context window. Shut down for free consumer accounts June 18, 2026.

autocomplete.sh

LLM-powered shell completion: hit Tab twice and it suggests the right command, using OpenAI, Groq, Anthropic, or a local Ollama model.

Void

Open-source, MIT-licensed VS Code fork positioned as a transparent alternative to Cursor, with agentic editing and support for local or hosted models.

Antigravity

Google's agent-first IDE built around Gemini 3, running parallel sub-agents that write code, configure routes, and test the UI they just built.

Cline

Open-source VS Code extension that turns the editor into an autonomous coding agent, using your own API key and an approval gate before each action.

Devin

Cognition AI's autonomous software engineer that plans, writes, debugs, and ships code with limited human oversight, across several model backends.

OpenHands

Open-source platform for autonomous coding agents built by All Hands AI, running in a sandboxed Docker environment to fix issues and open pull requests.

LangGraph

LangChain's graph-based runtime for stateful agents, modeling each step as a node with checkpointing at every super-step for time-travel debugging.