About
Last week I showed a roomful of GenAI enthusiasts some obvious and some not-so-obvious ways to, basically, make GitHub Copilot do their bidding.
My first point was that there are, like, a lot of AI-assisted development tools, some of them open and others not as open, most of them offering 🆓 plans1. So there’s really no excuse to avoid trying at least some of them out.
I also showed them a list of such tools, and I figured, why not share that list along with some of my notes. Please note that it’s by no means exhaustive, new tools seem to pop up daily. That being said, if you’re looking for something to start using, you might find something interesting below.
The awesome list
Dedicated IDEs
Cursor 🆓
A VS Code fork and one of the more interesting offerings in this space. Its concept of dynamic Rules that can auto-attached based on paths looks interesting and I wonder why more tools don’t follow suit. One of the IDEs I most want to dig into.
Windsurf 🆓
Also a VS Code fork, used to be cool – and maybe still is but has gone through some abrupt2 ownership changes lately.
Zed 🆓
My favorite agentic IDE at the moment. It can integrate with your GitHub Copilot subscription and models, but uses them with its own (vastly superior, imo), agentic implementation. Like, it’s not even a contest. Plus, it’s fast-fast. Like, Notepad++-fast for you Windows folks out there. Has pretty much replaced VS Code for me for lightweight code editing. It’s open-source. One of its creators worked on Atom. Love it.
IDE Extensions
Continue 🆓
Open-source VS Code & JetBrains extension, I’ve used this for the longest time, it was a lifesaver for specifying the exact context for the LLM at a time when GitHub Copilot would just automagically select 20 lines of context, no more no less. You can bring your own API key, and lets you configure your own Azure OpenAI deployments for added privacy and peace of mind. Its agent mode is a bit odd, and the UI a bit buggy, so I don’t use it that much anymore, especially with Copilot’s agent mode up and running.
Cline
VS Code open-source extension. You can bring your own API key, use Azure OpenAI, etc. It shows you a token and cost counter so you can see exactly how much money you’re spending to get that div centered just right. For whatever reason I don’t trust it (Cline, not its counter 🤷🏻♂️).
RooCode
A fork of Cline. Also open-source. Haven’t tried it yet.
KiloCode
This is <checks notes, coughs>, a Franken-merge of Cline, RooCode, and Continue 🥶. Apparently it’s not bad. Reminds me of this quote. Open-source
JetBrains Junie 🆓
Polished UI with some interesting features. A dealbreaker for me is that you can’t choose the model it uses per chat, you can only pick the default model between GPT-5, Sonnet 3.7 and Sonnet 4. You will need a JetBrains IDE license to use it.
Gemini Code Assist 🆓
I hadn’t realised this existed, but it looks cool. Available for VS Code and JetBrains. Haven’t tried it yet.
GitHub Copilot 🆓
My main driver. Has a wide variety of models, some very clear pricing and limits (i.e., one Sonnet 3.7 Thinking request is worth 1.25x “regular” requests. For comparison, Cursor will count Sonnet 3.7 Thinking to be worth 2x “regular” requests, and the list goes on). If you’re a subscriber you get GPT 4.1 for free with no limits. Has extensions for VS Code, Visual Studio, and JetBrains. Has a non-ideal agent mode, especially if you stack it against Zed – it’ll try to chunk files and look at them 50 lines at a time, and it’s just so annoying to see it fumble around trying to understand what it’s looking at. tl;dr: extensive model offering, very reasonable pricing, works in your IDE, poor agent mode but you can use Zed.
Command Line Agents
Aider
The OG CLI AI development tool. Open-source. You can connect it to pretty much anything, including Azure OpenAI. I’ve tried it a number of times, it looks great on paper, it has a very interesting architect mode where you can use an advanced but slow planner model, and a simpler but faster coder model to achieve world domination. It never clicked for me though, maybe because of its (lack of) integration with IDEs, maybe because I always needed to manually add context because it lacks an actual agent mode.
Claude Code
Open-source. Cool af. Only supports Claude but who cares that’s what we all use anyway. I’ve tried it once and it burned through tokens like a hot knife through butter. It has some interesting agentic features including writing and maintaining a todo list for each task. Anthropic encourages you to start multiple agents to go ahead and just do stuff, forget that code exists, etc.
Codex CLI
Also open-source. I…don’t know a lot about it except that it exists and it’s not as talked about as Claude Code 🤷🏻♂️.
Gemini CLI 🆓
I sometimes use Gemini 2.5 Pro as an alternative to Claude Sonnet and, when it doesn’t get stuck in thinking loops, it’s quite decent. You too can access it for pretty much free via the Gemini CLI, at the time of writing they offer 60 model requests per minute and 1,000 requests per day at no charge. Not sure how good the agent mode implementation is.
Cloud Agents
Or “Asynchronous Coding Agents” if you will. These are agents that will fetch your repositories, clone them to a Cloud VM, and work hard to do your bidding. All you need to do is to sit back and not think very hard about what could go wrong. I’m just a bit reluctant to do any serious work with them, since I usually hand them off small, targeted tasks and then it’s just faster to run them locally. They’re probably the future of development, but we’ll see.
Jules 🆓
I love its website. Supports 15 tasks/day for free at the time of writing. Haven’t tried it yet.
OpenAI Codex
Same as Jules, but from OpenAI. They use codex-1, a specialized model “fine-tuned to work in large codebases”. Haven’t tried it yet.
GitHub Copilot
It offers a cloud agent as well, that you can chat to and ask it to create PRs on your repos and all that. I’ve tried it a couple of times, and in some cases it worked really well, but since I generally want to be able to review and iterate on the results, I just find it faster to use an IDE.
Open SWE
An open cloud agent implementation that you can self-host on your own infrastructure. Looks interesting, and it’s definitely something worth looking at for bigger companies with a budget for running their own infra instead of paying other people to do it for them. That being said, I wonder if the code and docs are of the same quality as langchain’s 😉.