Show HN: p99.chat – the assistant for software performance optimization
p99.chat
Today, we're really excited to launch our new product: p99 (https://p99.chat), an assistant for software performance optimization.
Through CodSpeed, we have been working with hundreds of projects doing performance optimization. What struck us was how fragmented the tooling landscape is. You would identify a performance regression in their CI, or worse, in production, then disappear into a rabbit hole of benchmarking frameworks, sampling profilers, memory profilers, instrumentation tools, and visualization tools. With each tool coming with its own setup requirements, data formats, and workflows. This less interesting part of setup ends up taking most of the troubleshooting time.
Even for our own developer-focused tool, we worked hard on making the CI product as simple as possible for straightforward use cases, but the required integration level is so deep that there seems to be an inevitable high entry cost for performance tooling.
Our solution, p99 eliminates this friction with an integrated chat interface. Instead of context-switching between tools, you describe your performance challenge, share your code, and p99 orchestrates the entire analysis pipeline. It handles code analysis, instrumentation (using CodSpeed's CPU profiling), and visualization in one seamless conversation.
The goal is to cut setup overhead from hours to seconds, so you can focus on what matters: finding the issue and making your code faster.
Current features: Python and Rust execution, algorithm comparison, (differential) profiling, flamegraph and charts display.
Coming next: MCP integration for editor integration, additional instruments (memory profiling, walltime profiling, DB instrumentation, ...), GitHub integration for direct code analysis, and automated PR generation for performance improvements.
For now, the product is completely free. Although we require a login before creating chats to prevent abuse, you can check out a short demo here: https://youtu.be/yxY-nWNbK6M
Looking forward to your feedbacks and comments!