Leading Open Source alternatives to Copilot and Cursor
No-hype review of 7 OSS AI Code Assistants
7 Open Source AI Code Assistants
Over the past month, I've been exploring Open Source tools to help us developers write code faster and better. What started as a review of one tool expanded into testing several, as I found each one lacking. This post now features reviews of 7 different Open Source AI Code Assistants.
This experience made me realise, how much hype is there about Open Source AI Coding Assistants.
Whether it's from product owners, reporters, or social media influencers, grand promises are made about "innovative game-changers." In reality, these products are often just another VSCode fork or LLM wrapper with minimal added value. You get caught up in the excitement, feel the fear of missing out, and invest time trying the product—only to end up disappointed. This cycle can make you skeptical of new products in the Open Source AI Coding Assistant category, even those that might genuinely add value.
To evaluate their claims, I tested all these assistants myself. Here, I present my experience with 7 Open Source AI Coding Assistants without any hype or exaggeration.
Void editor - A VSCode fork with AI features and the most exaggeration. Easy to install and supports multiple LLMs directly configurable in extension settings. While promoted as an alternative to Cursor, it's an overstatement to call it that. It has limited features—only offering
Ctrl + Lfor chat and applying the suggested code with diff. This isn't enough added value to justify installing a VSCode fork; it could have been a VSCode extension in its current form. This is not truly a Cursor alternative at this point of time.Cody- A Copilot alternative using a WASM-based approach (running C++ code in the browser via Web Assembly). It supports multiple LLMs through a Sourcegraph server instance and has the most features among the products listed. However, I won’t use it, because: it's complex to install, requires a Sourcegraph account, and includes extensive telemetry.
Continue Dev- Another simpler Copilot alternative. It offers more features than Void and has simpler code than Cody, making it a good starting point for those wanting to create their own assistant. However, it still requires significant efforts. I couldn't get it to work—tried to build it from the source, no errors but the chat box didn't open. The lack of clear build instructions didn't help either. I will give more points to Void and Cody on this part. They were super easy to build from the source. I was planning to dig deeper and debug the installation issues but then I saw references to Continue account activation in the source code, that made me deprioritize investing more time on Continue for now. That said, it seems that considerable work has gone into building this product.
Tabby - AI code completion. I don’t want to talk much about it because only the client side is Open Source. The server side code has an ee (enterprise edition) license. So I’d thank the product owners to make some parts of their product Open Source but I won’t end up using the product in my day to day work because of the other essential part which is not Open Source. And hence no review for this one.
AI PR Reviewer by CodeRabbit - Disclaimer: I have been associated with this product. I love the amount of attention to details that has been given to the product and it does what it says without overpromising. The Open Source project is not maintained anymore though, and the Cloud product has moved way ahead of this Open Source project.
Plandex - Terminal-based code assistant agent using openai, multiple branches, rewind, accept/reject. Looks fun and seems to be the result of a lot of efforts to make the developer experience (DX) better within the terminal. But the terminal has its own limitations on DX when it comes to the engaging task of writing, reviewing, and interacting with the code in such a dynamic environment. I doubt it but I might keep trying this one for couple of weeks to see if I can find a workflow where this can become part of my daily routine, at least for some tasks. AGPL license.
Blinky - AI debugging agent. This is a vscode extension, provide the steps to reproduce and it will suggest next steps. I love the simplicity of this project (no vscode fork 🙂) and I might find the right configurations to use it in my workflow everyday. I haven’t got a chance to go through its code to understand what data it collects and how it deals with it. This might require a thorough review, if my concerns are addressed, I will write a detailed review on this one in couple of weeks.
If you have tried any of these tools, do share your experience. If you think I have rushed and made a bad judgement, I will be happy to look at the product again and adjust my opinions accordingly.
I am not at all against forking the projects or making the LLM wrappers. If it works for my use case, if it saves time for me, I don’t care how much effort was put into building the project, I’d use it, help improve it, and recommend it.
If you discovered an interesting Open-Source project and want me to feature it in the newsletter, get in touch with me.
To support this newsletter and Open-Source authors, follow #OpenSourceDiscovery on LinkedIn and Twitter