The right replacement depends on where you want the agent to live: in your terminal, inside your editor, or on a platform that runs tasks without you at the keyboard. Most of the agents here let you choose your own model, so the bigger differences are workflow and whether the project is still maintained under the name you know.
Why look elsewhere when Aider already pairs with you in the terminal and works with most large language models? It may be that you want a planning step before edits, the same agent inside your editor, or an agent that runs whole tasks on its own. The questions below take those in turn.
Which familiar names are missing, and why does it matter?
Three tools that often appear in lists like this aren't here, and each for a reason found on its own site. Continue's site says it was acquired by Cursor. Gemini CLI's page says its unpaid tier was replaced by Antigravity CLI in June. And the Roo Code domain now shows a cloud coding agent called Roomote. None of those is a judgement on quality; they've changed shape enough that they no longer answer the question of what to use instead of a terminal pair programmer.
That's the most useful thing to take from this page. Before you commit a team to any agent, open its own site and its repository and check two things: that it's still published under the name you expect, and that changes are still landing. If either answer is unclear, keep looking.
What are the eight agents, A to Z?
Names run alphabetically and sit in no particular order, because the right one depends on where you work. The line under each comes from the project itself.
Cline
Best for: Building an agent into your own products as well as daily use, since the vendor calls it one open source agent runtime.
Codex CLI
Best for: Automating repeatable work without leaving the terminal, as the vendor puts it, while it inspects, changes and runs code.
goose
Best for: Developers who also want help with research, writing, automation and data analysis, since the vendor calls it a general-purpose agent.
gptme
Best for: Working in your real environment; the vendor says it uses your shell, files, git and browser with any model.
OpenCode
Best for: Trying free models first, as the vendor says you can start with them or connect models from a long list of providers.
OpenHands
Best for: Starting locally and growing into a team setup; the vendor says you can use any model and scale with enterprise controls.
Plandex
Best for: Large coding tasks with many steps, which the vendor says it can plan and execute across dozens of files.
Qwen Code
Best for: Open source fans who want one agent in the terminal, editor, desktop, browser and chat, according to its own site.
Terminal, editor or a platform that runs the work?
Terminal agents. These feel closest to what you're leaving. Codex CLI, per its site, is a coding agent you run in your terminal against a local repository. gptme describes itself as an open source agent that lives in your terminal and works with several providers or a local model. Plandex calls itself a terminal-based AI development tool. Qwen Code says it's an open source coding agent. OpenCode, by its own description, runs from desktop, terminal and other surfaces. goose says it runs on your machine as a desktop app, CLI and API.
Editor and desktop agents. Cline describes Plan and Act modes, and says it runs in your editor, terminal or desktop. If you like the idea of reviewing a plan before any file changes, that split is the thing to try first.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Platforms that run tasks for you. OpenHands says it's an open source platform to build and run AI coding agents and automate engineering workflows. It's a different kind of tool: less a pair programmer, more a way to hand off a job and review the result.
| Tool | Kind of agent | Good for |
|---|---|---|
| Cline | Open source agent for editor and CLI | Developers who switch surfaces |
| Codex CLI | Terminal coding agent | Developers who want OpenAI's own agent |
| goose | Open source general AI agent | Running an agent locally |
| gptme | Terminal AI agent | Developers comfortable on the command line |
| OpenCode | Open source coding agent | Desktop and terminal users |
| OpenHands | Open source agent platform | Handing off whole tasks |
| Plandex | Terminal tool for large tasks | Big multi-file changes |
| Qwen Code | Open source terminal coding agent | Teams moving between tools during the day |
If you're also weighing up a hosted terminal agent, my Claude Code alternatives list covers that side, and my guide to OpenAI's coding agent walks through getting started with one.
Maintain an open source coding agent that belongs on this page? Send it in for a look and it may be added when the list is next checked.
What's my advice for choosing between them?
Run the same small, real task on two agents and compare the diffs, not the demos. Because most of these let you choose your own model, keep the model the same across both runs so you're judging the agent's workflow rather than the model behind it. Pick a task you understand well, such as adding a test for an existing function or renaming a field across a handful of files, so you can spot a wrong change at a glance.
Then think about where the agent will sit in your day. If you live in the terminal, stay with a terminal agent and the switch is smaller. If half your team uses an editor and half the command line, a tool that runs in both saves arguments. If you want to hand off whole tasks, a platform is worth the extra setup.
Signs a project is in good health before you rely on it:
- Recent commits and releases in its own repository, not just activity on a fork.
- A homepage and docs that use the same name as the package you install.
- Open issues that get replies, rather than a long list nobody has touched.
- Clear notes on which models and providers it supports, and how you add your own keys.
Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.
Frequently asked questions
Can I keep using my own API keys with these agents?
Most of the agents here say you can choose your own model. Check each project's docs for the providers it supports and whether it can run a local model.
Is a terminal agent safe to run on a work repository?
Treat it like any tool that can edit files and run commands. Work on a branch, review every diff before you commit, and check whether the agent asks before running shell commands.
What's the difference between a coding assistant and a coding agent?
An assistant suggests code as you type. An agent takes a task, plans the steps, edits files and can run commands or tests, then hands you the result to review.
Should a whole team use the same agent?
It helps for shared settings, rules files and reviews, but it isn't essential. Agree on how changes are reviewed and which models are allowed, then let people pick their surface.
Why did some well-known tools drop off this list?
Their own sites now describe an acquisition, a replacement product or a different hosted service, so they no longer fit a list of terminal and open source coding agents.
