Pomerado turns websites into MCP tools instead of browsing them
Pomerado launched on YC on Sunday with a one-line pitch: "Pomerado turns any website into an API / MCP, in minutes." It is in the YC Fall 2026 batch, and the core is open source under MIT on GitHub (about 190 stars, repo created October 5).
The target is the part of the web agents still have to click through: "legacy (insurance, vendor, logistics, payer) portals, EMRs, and government websites", sites that "only work through a browser, where computer use can be brittle." Instead of having an agent drive a browser every time, Pomerado generates a real integration once. You describe it in a single prompt and manage the process from Codex or Claude, with new APIs "in under 30 minutes." The hosted service handles login storage and automatic 2FA, and the "self healing APIs identify site changes and automatically kick off repairs."
The claimed numbers are 2 to 4 times faster, 10 to 1000 times cheaper, and 30% more reliable than browser use. Treat them as marketing. The launch post shows no method, and "10-1000x" is a range that covers almost anything. The architecture claim is the one that holds up on its own. A compiled integration that calls an endpoint is cheaper and more predictable than a model reasoning over screenshots on every run, because you pay for the reasoning once.
That fits a pattern from several directions this week. The Slovak corpus audit paper used an agent once to find problems and then froze the findings into tests that run in 0.7 seconds. Pomerado does the same for portals: browse once, compile, then call. The open questions are the obvious ones. What happens when the self-healing repair is wrong and nobody notices? And whose credentials are sitting in that "secure login storage"? For insurance and EMR portals, the second answer matters more than any speed number.
Launch: https://www.ycombinator.com/launches/WSd-pomerado-the-faster-cheaper-and-more-reliable-alternative-to-browser-use Repo: https://github.com/Pomerado/pomerado Site: https://pomerado.ai
← Back to all articles
The target is the part of the web agents still have to click through: "legacy (insurance, vendor, logistics, payer) portals, EMRs, and government websites", sites that "only work through a browser, where computer use can be brittle." Instead of having an agent drive a browser every time, Pomerado generates a real integration once. You describe it in a single prompt and manage the process from Codex or Claude, with new APIs "in under 30 minutes." The hosted service handles login storage and automatic 2FA, and the "self healing APIs identify site changes and automatically kick off repairs."
The claimed numbers are 2 to 4 times faster, 10 to 1000 times cheaper, and 30% more reliable than browser use. Treat them as marketing. The launch post shows no method, and "10-1000x" is a range that covers almost anything. The architecture claim is the one that holds up on its own. A compiled integration that calls an endpoint is cheaper and more predictable than a model reasoning over screenshots on every run, because you pay for the reasoning once.
That fits a pattern from several directions this week. The Slovak corpus audit paper used an agent once to find problems and then froze the findings into tests that run in 0.7 seconds. Pomerado does the same for portals: browse once, compile, then call. The open questions are the obvious ones. What happens when the self-healing repair is wrong and nobody notices? And whose credentials are sitting in that "secure login storage"? For insurance and EMR portals, the second answer matters more than any speed number.
Launch: https://www.ycombinator.com/launches/WSd-pomerado-the-faster-cheaper-and-more-reliable-alternative-to-browser-use Repo: https://github.com/Pomerado/pomerado Site: https://pomerado.ai
Comments