Launch HN: Pomerado (YC F26) – Turn any website into an MCP for agents
Pomerado (YC F26) has launched a platform to transform any website into a deterministic, self-healing API, tackling the significant hurdle of AI agents interacting with legacy systems. This innovative approach allows developers to create reliable 'Managed Control Planes' for agents in minutes, circumventing the slow and unreliable nature of traditional browser automation. By providing both open-source and hosted versions, Pomerado aims to unlock new capabilities for AI in diverse industries.
The Lowdown
Pomerado introduces a novel solution to a critical problem in the world of AI agents: their inability to reliably interact with real-world websites and legacy systems that lack proper APIs. Founded by ex-McKinsey and Scale AI professionals, Pomerado enables developers to quickly convert any website into a deterministic, self-healing API, effectively creating a 'Managed Control Plane' (MCP) for AI agents.
- The Problem: While AI agents are becoming increasingly intelligent, their utility is severely limited by their ability to interact with external systems. Many essential services, like airlines, banks, e-commerce, and government portals, lack modern APIs, forcing agents to rely on slow, expensive, and often unreliable browser automation, which also raises security and compliance concerns.
- The Genesis: The founders encountered this challenge firsthand while trying to enable their personal AI assistants to perform real-world tasks, such as booking flights, where agents consistently failed due to reliance on browser-based interactions.
- How Pomerado Works: The platform allows users (or coding agents) to provide a natural language description of a workflow. Pomerado's agent then explores the target website, performs the task once (e.g., via HTTP requests or Playwright script), and records the necessary data. This 'example run' is then used to create a generalizable and repeatable API. A 'Guardian' model reviews calls for guardrails, and synthetic tests ensure generalization.
- Key Features: Pomerado handles authentication and 2FA securely without reading user credentials, using deterministic sign-in processes and allowing users to link TOTP seeds or Pomerado phone numbers. Crucially, the system is self-healing: if an API breaks due to website changes, a maintenance job automatically kicks off to repair and improve the tool.
- Benefits: Unlike traditional computer vision or browser automation, Pomerado's MCPs are deterministic, leading to significantly faster and cheaper execution for repeated, high-volume tasks. The initial setup effort yields substantial long-term benefits.
- Availability & Use Cases: Pomerado offers both an open-source (MIT licensed) version and a free hosted platform. Early users include personal trainers automating client messaging, car dealerships integrating Grok Bots with legacy systems, and personal assistants interacting with a wide range of consumer websites.
In essence, Pomerado bridges the gap between sophisticated AI agents and the often-messy reality of web interactions, promising to make AI assistants far more practical and reliable for both personal and business workflows.