Reduck Raises $3M to Build the Fast Lane for AI Agents

French-American startup Reduck has raised $3 million in pre-seed funding to solve one of the most stubborn bottlenecks in agentic AI: most of the web still has no API for agents to call. The round was led by Daphni, with participation from Drysdale Ventures and Galion.exe, and the money will go toward engineering hires, infrastructure, and expanding the company's library of ready-to-run browser automation scripts, according to Tech.eu, which broke the news on October 22, 2026.

The problem Reduck is chasing

AI agents are being asked to do real work online: pull invoices from a dozen vendor portals, research prospects on LinkedIn, reconcile data across internal tools. But a large share of the websites and business applications agents need to touch were never built with programmatic access in mind. They expose no REST API, and many do not implement the Model Context Protocol interfaces that newer agent frameworks expect.

The fallback, as Reduck's founders frame it, is brute-force browsing: agents click, type, and scroll their way through interfaces designed for humans. That approach works, but it is slow, expensive in compute, and fragile — a redesigned page or an unexpected modal can derail an entire run. For teams deploying agents against insurance portals, healthcare systems, or internal customer relationship management tools, the brittleness is often the deal-breaker, not the model's intelligence. Even as frontier labs ship ever more capable agents — Anthropic's Claude Duet joined the proactive-agent race just this week — the plumbing problem remains: those agents still have to operate on the same brittle interfaces.

What Reduck MCP does

Reduck's answer is a platform it calls Reduck MCP: a hosted library of browser automation scripts that agents can discover, execute, and even create on demand. Instead of navigating a website step by step, an agent queries Reduck for an existing script matched to its goal — downloading invoices across multiple vendor portals is the company's go-to example — and runs it directly. If no script exists for a target site, the agent can generate one through Reduck's tooling, and the platform hosts and executes it going forward.

The company claims the scripted route makes agents roughly three times faster and three times cheaper on supported tasks. That figure comes from Reduck itself, as reported by Tech.eu, and applies to specific task classes rather than every possible workflow. The library currently holds more than 1,200 ready-to-run automations covering over 250 websites, and Reduck is developing its own agents to create and maintain scripts autonomously, aiming for a self-expanding automation layer that stays current as sites change.

The backers and the bet

Daphni led the round, joined by Drysdale Ventures and Galion.exe. The syndicate is an early-stage bet on a thesis gaining traction across the agent-infrastructure market: that the scarcest resource in agentic AI is not model capability but reliable execution tooling. The founders — chief executive Daniel Huynh, chief technology officer Louis Abraham, and chief operating officer Raphaël Millet — are positioning the company squarely in that tooling layer.

FinSMEs reports the funding will support growth across AI engineering, full-stack development, and infrastructure teams, alongside the build-out of a developer platform. That developer offering matters: it would let enterprise teams wire custom agent systems into applications that lack APIs, including specialized portals in insurance and healthcare. Shattered.io, analyzing the announcement, noted that independent benchmarking of agent task-completion claims is likely to become standard in vendor due diligence — a fair caveat on the 3x figures — but added that script-based execution through an MCP-style interface addresses a genuine adoption barrier.

What comes next

Reduck's roadmap has two notable branches. First, the autonomous script-maintenance system, which if it works would turn the automation library into something that heals itself as the web changes. Second, an exploration of generative engine optimization — using agents to automate knowledge work around how content surfaces in AI-driven search. Both are ambitious for a pre-seed company, but both map to the same insight: agents will spend most of their time on interfaces humans built for humans, and someone has to build the connective tissue.

The funding round lands amid a crowded week for agentic AI capital, from Tether's $15 million Series A in agentic-payments startup Mural to a Shanghai personal-agent company crossing $200 million in cumulative funding, with retailers like Walmart opening checkout flows to shopping agents in parallel. What distinguishes Reduck's raise is its size and specificity: a small, focused bet that the agent economy's real unlock is not smarter models but boring, reliable plumbing.

Sources: Tech.eu; FinSMEs; Shattered.io