AI startup Reduck has raised $3 million in Pre-Seed funding to develop browser automation infrastructure that allows AI agents to interact with websites and applications without dedicated APIs. The round was led by Daphni, with participation from Drysdale Ventures and Galion.exe. The company is building Reduck MCP, a platform that enables AI agents to discover, execute and create reusable browser automation scripts, reducing the need to navigate websites manually during each task.
Founded by Daniel Huynh, CEO, Louis Abraham, CTO, and Raphaël Millet, COO, Reduck focuses on improving how AI agents interact with online services that lack programmatic access. Its technology is designed to address the speed, cost and reliability limitations of conventional browser-based agents, particularly when handling repetitive workflows across multiple websites.
Building Browser Automation Infrastructure for AI Agents
AI agents increasingly use browsers to interact with websites and applications that do not provide APIs or Model Context Protocol (MCP) integrations. While this approach allows agents to perform tasks such as collecting information, downloading documents and updating records, it often requires repeated navigation, screenshots and model interactions. These processes can be slow and expensive, particularly when agents must complete complex workflows or operate across multiple services.
Reduck addresses these limitations through Reduck MCP, which allows agents to use reusable browser automation scripts instead of navigating each website from scratch. The platform enables agents to discover existing scripts, execute them as tools and create new automations when needed. According to the company, this approach makes agents approximately three times faster and less expensive across a range of benchmarked tasks, including retrieving invoices from multiple vendors and conducting research through LinkedIn.
The platform currently provides more than 1,200 ready-to-run scripts covering over 250 websites. These scripts can support workflows such as downloading SaaS invoices, extracting sales prospect information and collecting data for enterprise demonstrations. Reduck reports that customers have reduced some recurring administrative tasks from several hours to minutes by using its automation technology.
The company is also developing an autonomous system of AI agents that can generate and repair automation scripts at scale. This is intended to improve reliability as websites change and reduce the manual maintenance required to keep browser automations operational.
Funding to Expand Automation Capabilities and Developer Tools
Reduck was founded by three entrepreneurs with backgrounds in AI research, cybersecurity and browser automation. CEO Daniel Huynh and COO Raphaël Millet previously co-founded Mithril Security, which was acquired by H Company. Before the acquisition, Mithril Security developed LaVague, an open-source AI web agent, and worked on privacy-focused language model deployment. CTO Louis Abraham previously co-founded encryption platform Secrecy and worked on AI agent research and production systems at H Company, where he met Huynh and Millet.
Following the funding round, Reduck has expanded its team to eight employees across AI engineering, full-stack development and infrastructure. The company plans to continue improving the reliability and coverage of its automation library while developing tools that allow AI agents to interact with a broader range of online services and enterprise applications.
Reduck is also preparing a developer offering that will allow teams to build custom AI agent systems connected to environments without APIs, including specialised insurance and healthcare portals and internal CRM systems. Alongside this, the company is exploring vertically integrated applications built on its automation infrastructure, initially focusing on generative engine optimisation (GEO).
With the new funding, Reduck aims to expand the infrastructure supporting AI agents as businesses look to automate workflows across websites and internal systems that are not designed for direct AI integration.
