Agents that write, run and fix code in your terminal or editor.
Updated
Anthropic's coding agent that reads your codebase, edits files and runs commands from the terminal, IDEs, a desktop app or the browser. It needs a Claude subscription or an API key.
OpenAI's open-source coding agent for the terminal, released under Apache-2.0, which you use with a ChatGPT plan or an API key.
Google's open-source terminal agent built on Gemini models, released under Apache-2.0. A personal Google account allows free use within limits.
Code editor based on VS Code with an AI agent, chat and tab completion woven into everyday editing.
Formerly Windsurf, this AI code editor is now Devin Desktop from Cognition, the company behind the Devin agent, built for agent-driven coding.
Open-source coding agent that lives in VS Code and other editors as an extension, working with your own API key or a local model.
Open-source pair programmer for the terminal that edits files in a git repository and commits each change, compatible with many cloud and local models.
Open-source platform, formerly OpenDevin, for autonomous coding agents that work in a sandbox and start from a local interface, a CLI or the cloud.
GitHub's assistant for editors and github.com: completions, chat and a coding agent that opens pull requests. A free plan with limits is available.
Open-source Python framework for orchestrating teams of AI agents, each with its own role and tasks, when one agent is not enough.
A coding agent goes further than autocomplete. It reads your project, plans a change, edits several files, runs commands and tests, and fixes what fails — with you approving the steps or reviewing the result. This page lists agents in their different forms: terminal agents such as Claude Code, Codex CLI, Gemini CLI and Aider; editors with agents built in, such as Cursor and Devin Desktop; editor extensions such as Cline and GitHub Copilot; and frameworks for building your own agents, OpenHands and CrewAI.
First decide where you want to work. If you live in the terminal, a CLI agent fits; if you prefer to stay in the editor, pick an editor or an extension. Then look at the model behind it, because an agent is only as capable as the model it uses: the coding, web dev and agents leaderboards help here. Some agents are tied to one vendor's models, others let you bring any API key or even a local model.
An agent executes commands on your machine and changes your files. Work in a version-controlled repository, review what it proposes before accepting, and be careful with permissions that let it run anything without asking. Agents can also load skills — reusable instructions and scripts — which you will find in the skills directory.
Plans and the usage they include change often. The card shows what the vendor page said on the date of our check; the vendor's page is always the final word.