Cursor Desktop editor, plus CLI
An AI code editor built around agents you hand tasks to. Its docs cover Agent mode, Rules, Skills, MCP servers and a CLI.
Our take, best for: Builders who want one editor where agents do most of the typing.
Guide · Last reviewed
AI coding agents for builders: what each one is good at
Coding agents now live in three places: inside your editor, in your terminal, and in the cloud working on your repo while you do something else. Most builders end up using one from each group. Below is what each tool is (from its own docs), plus our take on who it suits. We left out pricing because it changes monthly; check each link before you pay.
At a glance
| Tool | Where it runs | Our take: best for |
|---|---|---|
| Cursor | Desktop editor, plus CLI | Builders who want one editor where agents do most of the typing. |
| GitHub Copilot | Editor, terminal, desktop and GitHub.com | Teams already living in GitHub issues and PRs. |
| Devin Desktop (formerly Windsurf) | Desktop IDE plus cloud agents | Windsurf users, and anyone who wants local and cloud agents in one window. |
| Kiro | Agentic IDE (from AWS) | Builders who like to plan with a written spec before code gets generated. |
| Cline | Open source (Apache-2.0): IDE extension, CLI or SDK | People who want an open-source agent and full control over which model they pay for. |
| Claude Code | Terminal and IDE (Anthropic) | Long, multi-step changes in an existing codebase. |
| Codex | Open-source CLI (Apache-2.0), IDE extension, desktop app, cloud | ChatGPT subscribers, and anyone who wants to run agents in CI pipelines. |
| Gemini CLI | Open source (Apache-2.0) terminal agent | Trying a terminal agent with Google's models, or forking one. |
| OpenCode | Open source (MIT) terminal agent | Open-source fans who want a model-agnostic terminal agent. |
| Aider | Open source (Apache-2.0) terminal pair programmer | Small, focused edits with a clean git history. |
| Amp | CLI and cloud runner | Builders who want to hand off long tasks and check back later. |
| Jules | Cloud agent (Google) connected to GitHub | Clearing small issues and chores without tying up your laptop. |
| Devin | Cloud agents (Cognition) | Teams with a big backlog of well-scoped tickets. |
| Codex cloud and Copilot on GitHub | Cloud | Whoever already pays for ChatGPT or Copilot: try these before buying anything new. |
Keep this list current without the research
Tools change every week. Hotfix Daily tells you what changed, why it matters and links the source. Free, weekdays, about 3 minutes.
In your editor (IDE agents)
Best when you want to watch and steer changes as they happen.
GitHub Copilot Editor, terminal, desktop and GitHub.com
GitHub's coding agent works from an open issue through to reviewing and merging a pull request, and runs in your editor, the terminal, a desktop app and on GitHub itself.
Our take, best for: Teams already living in GitHub issues and PRs.
Devin Desktop (formerly Windsurf) Desktop IDE plus cloud agents
Cognition renamed Windsurf to Devin Desktop on June 2, 2026. It keeps the full IDE and adds an Agent Command Center for managing local and cloud agents from one place.
Our take, best for: Windsurf users, and anyone who wants local and cloud agents in one window.
Kiro Agentic IDE (from AWS)
Kiro turns prompts into executable specs, checks code correctness, and runs parallel agents across large codebases. It is built by Amazon Web Services.
Our take, best for: Builders who like to plan with a written spec before code gets generated.
Cline Open source (Apache-2.0): IDE extension, CLI or SDK
An open-source coding agent with Plan and Act modes and MCP support. You bring your own model provider.
Our take, best for: People who want an open-source agent and full control over which model they pay for.
In your terminal (CLI agents)
Best for real repos, scripts and automation. They read your codebase, edit files and run commands.
Claude Code Terminal and IDE (Anthropic)
Anthropic's agentic coding tool understands your codebase, edits files and runs commands. It reads project instructions from CLAUDE.md or AGENTS.md files.
Our take, best for: Long, multi-step changes in an existing codebase.
Codex Open-source CLI (Apache-2.0), IDE extension, desktop app, cloud
OpenAI's coding agent runs locally in your terminal and can be scripted in CI with codex exec. There is also an IDE extension, a desktop app (macOS since Feb 2, 2026; Windows since Mar 4, 2026) and a cloud version. Usage is included in paid ChatGPT plans.
Our take, best for: ChatGPT subscribers, and anyone who wants to run agents in CI pipelines.
Gemini CLI Open source (Apache-2.0) terminal agent
An open-source AI agent that brings Google's Gemini models into your terminal.
Our take, best for: Trying a terminal agent with Google's models, or forking one.
OpenCode Open source (MIT) terminal agent
An open-source coding agent you can point at the model provider of your choice.
Our take, best for: Open-source fans who want a model-agnostic terminal agent.
Aider Open source (Apache-2.0) terminal pair programmer
AI pair programming in your terminal, working directly on your git repo.
Our take, best for: Small, focused edits with a clean git history.
Amp CLI and cloud runner
A coding agent and development environment. You can run it on your own machine with its CLI, or unattended in the cloud.
Our take, best for: Builders who want to hand off long tasks and check back later.
In the cloud (async agents)
Best for backlog chores: give it an issue, get a pull request back.
Jules Cloud agent (Google) connected to GitHub
Clones your repo into a cloud VM, plans and makes the change, runs tests and can open a pull request. You can trigger it from GitHub or a GitHub Action.
Our take, best for: Clearing small issues and chores without tying up your laptop.
Devin Cloud agents (Cognition)
An autonomous AI software engineer that runs parallel cloud agents, aimed at engineering teams.
Our take, best for: Teams with a big backlog of well-scoped tickets.
Codex cloud and Copilot on GitHub Cloud
Both OpenAI (chatgpt.com/codex) and GitHub let you delegate a task in the browser and review the result as a diff or PR.
Our take, best for: Whoever already pays for ChatGPT or Copilot: try these before buying anything new.
How to choose (our take)
- Just starting? Pick one editor agent (Cursor or Copilot) and learn to review its diffs. That habit matters more than which tool you pick.
- Already comfortable in a terminal? Add a CLI agent (Claude Code, Codex or Gemini CLI). They shine on multi-file changes and can run your tests themselves.
- Want open source or your own models? Cline, OpenCode, Aider, Codex CLI and Gemini CLI all publish their code on GitHub.
- Have a backlog? Try a cloud agent (Jules, Copilot on GitHub, Codex cloud) on low-risk issues first.
Four habits that make any agent better
- Give every agent the same instructions file. AGENTS.md is an open format, a README for agents. Claude Code also reads CLAUDE.md. Put your build, test and style rules there once. Source
- Checkpoint with git. OpenAI's Codex docs recommend creating git checkpoints before and after a task so you can roll back. Source
- Make it show proof. Ask the agent to run tests or the build before it says "done", then have a fresh session review the diff. Anthropic says this is the first layer of its own code review process. Source
- Connect tools with MCP. The Model Context Protocol is the common way to plug databases, docs and apps into agents. Cursor, Cline, Unity and Unreal all support it. Source
FAQ
What is an AI coding agent?
A tool that doesn't just suggest code: it reads your project, makes changes across files, runs commands like tests, and reports back. You review and approve the result.
Which AI coding agent is best?
There isn't one winner. Editor agents suit hands-on work, terminal agents suit larger changes and automation, and cloud agents suit backlog chores. Most builders use one from two of these groups.
Are there free or open-source coding agents?
Yes. Cline, OpenCode, Aider, Gemini CLI and OpenAI's Codex CLI publish their source code on GitHub under open-source licenses. You still pay for model usage unless you run a local model.
What happened to Windsurf?
Cognition renamed Windsurf to Devin Desktop on June 2, 2026. The IDE, plans and settings carried over.
Related
- Vibe coding tools: build apps by describing them (October 2026)
- AI game dev tools in 2026: engines, assets, NPCs and prototyping
- The AI Builder Starter Stack (free PDF)
- Daily issues archive
How we write these: each tool description is based on the tool's own site or docs (linked as Source) and checked on the date above. "Our take" lines are our opinion. Something out of date? Email hello@hotfixdaily.com. No tool paid to be listed.
Get the next issue in your inbox
Five things worth your attention in AI, agents, vibe coding and game dev, every weekday. Plus the free AI Builder Starter Stack PDF when you join.