TensorZero
lastmod 2025-06-10created 2025-02-272 edits
creates a feedback loop for optimizing LLM applications — turning production data into smarter, faster, and cheaper models.
See also
Favorite site
Articles
- Reverse engineering Cursor's LLM client | GeekNews
- [Original] Reverse Engineering Cursor's LLM Client · TensorZero
- Shares the experience of using TensorZero as an open-source proxy to intercept and analyze traffic between Cursor and LLM providers (OpenAI, etc.), observing prompts, models, and inference results in real time and running optimization experiments
- Cursor lets you override the Base URL and model name for LLM calls, so your own proxy (TensorZero) can be integrated easily
- Internally, Cursor calls the LLM through its own servers, so a full proxy setup requires Ngrok + an Nginx reverse proxy and CORS header configuration
- Through the proxy you can observe everything Cursor actually sends to the LLM, including the system prompt, user prompt, and inline code edit requests, and you can switch and experiment between different LLMs (A/B tests) in real time
- Analysis of Cursor's system prompt shows that with a prompt of only about 642 tokens, the LLM understands and handles most software engineering context. Code edits are handled by a separate "apply model" (a less intelligent auxiliary model)
- A proxy setup like TensorZero enables per-user LLM experiments and feedback-based optimization; this setup is ideal for quality evaluation of coding assistants (A/B tests), prompt optimization, and real-world usage monitoring