Subscribe
Sign in
CursorLens — Open Source dashboard and analytics for Cursor IDE
An open-source dashboard for Cursor.sh IDE. Log AI code generations, track usage, and control AI models (including local ones). Run locally or use upcoming hosted version.
56
Replies
@hamedmp congrats on the launch. Pretty excited to see how it is going to make coding workflows better! 🚀
Hi @sentry_co just a few hours ago we shipped Ollama support. If you download the model with Ollama and set it in configuration, you can use it. https://x.com/cursorlens/status/...
tbh I had this thing on my mind to try various LLMs locally, I might actually have a usecase for it. Looks intresting enough to me
This sounds interesting! How does Cursor Lens handle privacy when tracking AI usage? Curious about the specifics! @hamedmp
Happy to hear that @ash_barbour, let me know what more we should add to make it even easier
You're launching an open-source dashboard for Cursor.sh IDE, helping developers monitor and optimize their AI coding workflow in real-time. With Cursor Lens, users can track AI usage, estimate costs, and customize model configurations. Keep driving this innovation forward, empowering developers to make the most of AI tools. 🚀🖥️✨
Thanks @zane12580 for exploring CursorLens, let us know your feedback and please share with your engineer friends ❤️
Congrats on the launch of Cursor Lens, @hamedmp! This is exactly what the developer community needs right now, especially with the growing reliance on AI for coding. The real-time analytics feature is a game-changer, providing insights that can help us optimize our coding workflows effectively. The fact that you support multiple AI models is also a big plus. I'm particularly excited about the customizable configurations—this will definitely allow for more tailored interactions. Can't wait to dive into it and see how it enhances my coding efficiency. Upvoted! 🚀
Thanks @star_boat for the feedback. As far as all the logs are gathered and saved, many tasks can be done on top.
Starting from:
1. Number of prompts, tokens, cost
2. Configuration of which model to use (technically any model can be used now)
3. It can also be configured to retry queries if an agent LLM sees the 1st LLM has failed, so send it to a more powerful or different LLM.
I put a feedback form on the website to gather ideas from the community as well, please let me know what can impact your work significantly.
Thanks again for checking this out <3
This is an interesting launch, @hamedmp! I'm curious about the decision to include support for multiple AI providers. Are there specific features or capabilities that set Cursor Lens apart from other tools in terms of integration with these providers? Also, for teams trying to manage costs, how does the token usage tracking work in practice? Is there a way to visualize trends over time? The analytics aspect sounds like it could be a game-changer for devs. Would love to hear more about the future roadmap and any potential integrations you have in mind! 🧐
Thanks for the feedback @antoine_gauthier_ great points!
At the moment we are using the AI SDK, but nothing stops us from integrating with any provider, including local ones. You can the upcoming providers and the roadmap over here: https://www.cursorlens.com/docs/...
The visualization of tokens used over time is already on the Stats page but below is where I put the screenshot. The Stat page can be like a Posthog dashboard where you mix and match what you need. As far as there is data, everything else is just a SQL query. Totally doable!
Let me know what you think about the roadmap and I'd love to hear more about your ideas
🙇🏻
Congrats on launching Cursor Lens, @hamedmp! 🚀 The gap you’re addressing is super relevant—AI tools can be a bit of a black box sometimes. Love the idea of offering real-time analytics, especially for teams managing costs and efficiency.
I'm particularly curious about the customizable AI model configurations. That sounds like a game-changer for tailoring interactions! Do you have any MVPs in mind for future features?
Also, any tips for new users on getting the most out of the analytics section? Looking forward to seeing how this evolves! Keep up the great work and happy shipping!
Thank you @robertthomas2 for your kind feedback! I share the same belief and think we'll have many AI Code agents in the future, a central hub to control them outside of the box can be helpful, both for programmers and also for Code agents as they can offload some of the basics to CursorLens and focus on better diff applier and multi-file editor, ...
Overall, I see these features as low-hanging fruit (can be done by the end of next week):
1. Number of prompts, tokens are done, cost to be added
2. Configuration of model to use (technically any model can be used now)
3. Adding retry feature: an agent LLM can check if the 1st LLM has failed, so send it to a more powerful or different LLM.
Keep your ideas coming, these are just on top of the head ones, but I think there opportunities are endless.
This looks promising! I love the idea of real-time analytics for AI usage. @hamedmp, have you considered integrating more AI providers in future updates? It would be great to see broader support for different models to enhance compatibility. How does the token usage tracking work exactly?
Hi @christopherdavidanderson,
Thank you for the feedback. I think it's outlined in the upcoming work on the roadmap: https://www.cursorlens.com/docs/...
How the tracking we use Vercel AI SDK which provides us the total talking usage over here: https://sdk.vercel.ai/docs/ai-sd...
Yess, this is it @christopherdavidanderson thanks for sharing these @stephen_rayner 🤗
This looks really interesting! I'm curious about the real-time analytics feature you mentioned. How specific can the data get regarding AI interactions? For instance, will it provide insights on which coding patterns are most frequently assisted by AI? Also, could you elaborate on how customizable the AI model configurations are? Looking forward to trying Cursor Lens out!
Hi @alexanderscottwilliams, thanks for the feedback and the questions.
Coding patterns tie closely with what @_ivan1 mentioned as ML capabilities. A path to explore more. I'm willing to hear your ideas on this.
Re. interactions, we have access to full requests being sent and received by Cursor, so except inside Cursor actions (like if the response is accepted or not, ...) we can have good intel on the interactions of the developer with the LLM models.
On LLM configuration, if you mean which LLMs we can support, atm we are using AI SDK by Vercel which lets us customize all major parameters on an API call, even Anthropic's prompt caching is now added and we can do prompt caching to reduce costs for LLM calls.
Looks like a really useful tool for devs, love that it's open-source and really appreciate the focus on optimizing AI-assisted coding, could be a game-changer for many teams. Any plans to add integrations with other dev tools like GitHub or JIRA? Would be awesome to track AI usage across different stages of the dev process. Keep up the great work!
Cursor Lens