Pomerado is a platform for turning websites without APIs into tools that AI agents can use through MCP. Its founders argue that agents are often limited less by their reasoning than by their ability to interact with systems such as airline, banking, insurance, government, and legacy business portals. Browser automation can be slow and unreliable, especially for repeated workflows, and some organizations prohibit it for security or compliance reasons. Pomerado aims to make these connections self-serve: a user describes a workflow in natural language, answers follow-up questions, and receives a reusable integration, typically in under an hour.
To build one, an agent first explores a site and completes the task, recording how it works and producing HTTP requests or a Playwright script. The platform then generalizes that run into a deterministic API, tests it with synthetic inputs, checks the code against guardrails, and publishes an input/output schema. It provides a separate sign-in process intended to keep the model and recorders from seeing credentials, and can support two-factor authentication. If an integration later breaks or needs new functionality, a maintenance process can repair or extend it. The founders say later calls usually take 5-30 seconds, making the approach useful for repeated, high-volume tasks. Examples include a trainer automating client communications, car dealerships connecting agents to legacy systems, and personal assistants using consumer websites. Pomerado offers both an MIT-licensed open-source version and a hosted service.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.