Skip to content
LATEST
Technology

Pomerado targets browser agents with $0.03 task claim

The YC-backed startup says its hosted MCP integrations are faster and cheaper than browser automation for repeated website workflows.

Pomerado targets browser agents with $0.03 task claim

Pomerado, a Y Combinator-backed startup, launched on Hacker News on Nov. 11 with a tool it says can turn websites into API/MCP integrations in minutes.

The company said a sample employee-onboarding workflow took 8 minutes 16 seconds and cost $0.03 using Pomerado, compared with 21 minutes 22 seconds and $41.01 for Codex Computer Use.

Pomerado’s pitch is aimed at a bottleneck in agent software: many airlines, banks, e-commerce sites, insurance portals, government services and internal tools do not offer application programming interfaces, or APIs. MCP refers to Model Context Protocol, a standard for connecting AI agents to external tools and data sources.

On its website, Pomerado says its product creates deterministic APIs for websites, meaning the integration follows a repeatable workflow at runtime rather than asking a browser agent to reason through each page again. “Pomerado turns any website into an API/MCP in minutes,” the company said on its site.

The startup said users can begin with a natural-language description, answer questions about the task and receive an MCP in less than an hour. Future calls to the generated tool usually return in 5 seconds to 30 seconds, according to the company’s Hacker News post.

Pomerado said its system first explores a site and performs a task using HTTP requests or a Playwright script, then uses that run to build a reusable integration. It also said a separate Guardian model checks calls against guardrails, while tests using synthetic inputs are intended to reduce overfitting to one example.

The company is positioning the product against browser automation and computer-use agents, which operate websites visually and step by step. Pomerado said computer use can be slow, costly and unreliable, and said many companies have banned it because of security and compliance concerns.

The startup also claims to handle authentication and two-factor authentication. It said sign-in is passed to a deterministic process so its models and recorders do not read customer credentials, and users can store TOTP seeds or link an account to a Pomerado phone number for autofill.

Pomerado’s website presents a benchmark across 30 tasks in 25 business applications, with one try each. It says Pomerado was 2 times faster, 10 to 1,000 times cheaper and 30% more accurate than Astra Computer Use, Instinct and Muse for browser use; those figures are company claims and were not independently verified in the source material.

The company said its integrations are hosted, scaled and access controlled, so customers do not deploy infrastructure. Its website shows examples spanning flight search, order history, hotel search, reservations, procurement portals, HR systems, billing systems and claims systems.

The main uncertainty is durability outside the company’s examples.

The launch also highlights a broader divide in agent infrastructure: whether AI systems should keep operating software through browsers, or whether vendors can translate recurring website tasks into more stable APIs. Pomerado’s own framing is that repeated, high-volume workflows benefit most from taking longer on the first run.

Source: Hacker News launch post and Pomerado website, Nov. 11.

More stories

Latest news