AI agents are good at clicking. They are slow and expensive at doing it across a hundred invoices.
Reduck, split between France and the United States, raised $3M in a pre-seed led by Daphni, with Drysdale Ventures and Galion.exe joining. Founders Daniel Huynh, Louis Abraham and Raphaël Millet are targeting the gap left by sites and internal tools that offer neither an API nor a Model Context Protocol interface.
Scripts instead of step-by-step clicks
Its platform, Reduck MCP, lets agents discover and run prebuilt browser automation scripts rather than navigating each page manually. The company says that makes tasks roughly three times faster and cheaper, and its library already holds more than 1,200 scripts across 250-plus websites.
Invoices from multiple vendors and LinkedIn research are the early examples. Reduck is also building agents that write and maintain those scripts themselves, so the library grows without manual engineering.
A developer offering is in the works for teams connecting agents to portals that never shipped an API, with insurance and healthcare listed alongside internal CRM systems, and generative engine optimisation floated as a first knowledge-work use case.
The bet is that most of the enterprise web will never be API-ready, leaving browser automation as the durable bridge that agents depend on.
