Cursor Lens

Dashboard and analytics for Cursor IDE

240 followers

Dashboard and analytics for Cursor IDE

240 followers

Visit website

AI Coding Agents

•

AI Infrastructure Tools

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.

Free

Launch tags:Open Source•Software Engineering•Developer Tools

Launch Team / Built With

Webflow | AI site builder — Start fast. Build right.

Start fast. Build right.

Promoted

Cursor Lens

Maker

📌

🚀 Excited to launch Cursor Lens on Product Hunt today! As developers increasingly rely on AI-assisted coding, we realized there was a crucial gap: understanding how we interact with these AI tools. That's why we created Cursor Lens - to shed light on your AI-coding workflow and help you optimize it. 🔍 Key highlights: Real-time analytics on your AI usage Support for multiple AI providers (OpenAI, Anthropic, and more) Token usage tracking and cost estimation Customizable AI model configurations 🌟 Perfect for: Developers looking to improve their AI-assisted coding efficiency Teams wanting to optimize their AI usage and costs Anyone curious about their interaction patterns with AI coding assistants We're open-source and built with Next.js, PostgreSQL, and Vercel AI SDK. Try it out and let us know what you think! We're eager to hear your feedback and ideas for future features. Happy coding, and may your AI assistants be ever insightful! 🖥️✨

Report

1yr ago

Fotographer AI

@hamedmp congrats on the launch. Pretty excited to see how it is going to make coding workflows better! 🚀

Report

1yr ago

DiffSense

Very cool! I use cursor everyday! How does the local models integrate? I use my own openAI key. but I didn't know it supported local models. Care to elaborate?

Report

1yr ago

Cursor Lens

Maker

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/...

Report

1yr ago

This sounds really promising! @hamedmp, I'm curious about the real-time analytics feature. How detailed is the tracking? Can we get insights into specific tasks or just overall usage? Would love to understand how this could impact my workflow!

Report

1yr ago

Cursor Lens

Maker

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

Report

1yr ago

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?

Report

1yr ago

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...

Report

1yr ago

Cursor Lens

Maker

Yess, this is it @christopherdavidanderson thanks for sharing these @stephen_rayner 🤗

Report

1yr ago

I haven't been using Cursor for long, but I've already become an annual paying user. I'm not sure yet if I need CursorLens. However, I definitely have to support open-source!

Report

1yr ago

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!

Report

1yr ago

Cursor Lens

Maker

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.

Report

1yr ago

TTSynth.com

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!

Report

1yr ago

Cursor Lens

Maker

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.

Report

1yr ago

1 2 3

•••