Launching today
Pulse
See which MCP tools get used, fail, and slow down
1 follower
See which MCP tools get used, fail, and slow down
1 follower
Pulse shows usage, errors, and latency for your MCP tools without collecting prompts, arguments, or results. Use the open-source SDK locally or connect to Pulse Cloud for a shared dashboard. Try the public demo without signing up.
Hey Product Hunt! I'm Selim, the maker of Pulse 👋
Once an MCP server is running, the next questions are practical: which tools are getting used, which calls fail, and where are handlers slowing down? Pulse brings those signals into one dashboard without collecting prompts, tool arguments, or results.
You can explore tool usage, handler outcomes, p50/p95 duration, and release comparisons. The SDK is MIT licensed and can run locally; Pulse Cloud is an optional managed destination with a shared dashboard. The Cloud interface supports English and Turkish.
Try the interactive demo without signing up: https://pulse.reviseflow.io/en/demo. Its sample data is clearly labelled. The public SDK and setup guides are linked on this launch.
We built Pulse with GPT-6 Astra in Codex throughout the project: the independent MIT SDK, managed Cloud service, dashboard, tests, and documentation. The challenge was turning an idea into a working product with clear privacy boundaries and trustworthy measurements. Thanks to ChatGPT and OpenAI for the tools that helped us build it.
If you build MCP servers, which would help you most: finding slow tools, investigating failures, or comparing releases? I'd love your feedback on the workflow and the signals you need.