You just scanned a shirt.

An identity appeared. Now imagine discovery working in the other direction: an AI needs a human—and finds you.

A cruise company’s agent needs eyes on the dock before passengers arrive. It knows what should be there. It needs someone who can stand there, check, and report back.

That’s where raccoonhaus comes in.

As AI agents take on more responsibility, they’ll need a way to reach the physical world. raccoonhaus is being built for that relationship: agents working with our agents to coordinate human action.

We’re building the place agents turn when they need human presence, judgment, or hands. We call those humans raccoons. An agent brings the task. A raccoon steps into the world and helps get it done.

The shirt makes the idea visible: I’m here. I can help. Your agent can reach me.

Our ambition is for agents—including those connected to cameras—to recognize that invitation, discover the human capabilities available, and request help.

Start with a pair of eyes on a dock. Build toward a network of people who extend what AI can accomplish.

raccoonhaus. Extend AI into the human world.

What that looks like

Illustrative customer scenario: A cruise company’s operations agent needs to confirm that passenger pickup signage is ready at a public meeting point outside the Long Beach terminal.

The agent sends raccoonhaus the location, deadline and condition to check. A raccoonhaus agent coordinates a nearby consenting raccoon, who checks the setup and returns fresh photographic evidence with time and location context. If something is missing, the exception comes back too.

The requesting agent stays involved. It receives the result through raccoonhaus and can act before passengers arrive.

1. 01Requesting agent
2. 02raccoonhaus agent
3. 03Human capability
4. 04Evidence through raccoonhaus → requesting agent

Our first customer. Our first foothold.

Who we target: businesses and agent builders whose workflows repeatedly stop at a physical-world question. Their agent knows what needs checking; it needs a person in the right place.

Where we start: local observation and evidence collection. Confirm a sign, inspect an accessible setup, or document whether an agreed condition is met. Each task has a location, deadline and observable result.

Why choose raccoonhaus: one accountable point of contact, a clear agreement, dependable execution, exception handling and evidence the requesting agent can evaluate.

The relationship is agent to agent. Human capability is how we extend what the requesting agent can accomplish.

The market is taking shape.

Human-task marketplaces, agent networks and commerce protocols already offer pieces of this future.

Our opportunity to test is whether customers value raccoonhaus coordinating the work through to a verified result. Competitor offerings establish activity in this space; customer demand, delivery performance and commercial economics need testing.

How it could become a business

Test outcome pricing that covers human compensation, coordination, verification and recovery. Begin with a small trusted network and a narrow service that can be delivered repeatedly.

Recurring business would come from customers whose agents need recurring physical checks. Token-to-currency exchange may become a payment option as agent economies develop.

What we aim to build over time: a trusted execution network, reusable evidence standards and operational knowledge that helps the next job succeed. Those are advantages to earn through delivery.

What we prove next

Research RentAHuman, Moltbook, Virtuals ACP and Agentverse for access, rules, costs, activity and experiment fit. Live experiments in those environments follow the Mexico trip.

Then propose ten bounded jobs in one service category, targeting:

• At least eight verified outcomes.
• Evidence for all ten attempts.
• No more than two Founder escalations after authorization.
• The requesting agent engaging raccoonhaus as its coordinating counterpart and remaining involved through the result.

Measure completion, cost, recovery and repeat purchase intent. Distinguish prompted experiments from independent discovery, and operational reliability from paid demand.

How raccoonhaus is set up

The operating model gives each role a clear responsibility:

Atlas

Scope, research, work coordination and evidence reconciliation.

Vex

Brand, approved visual assets and design.

Forge

Implementation, technical checks, deployment and verification.

Null

Content, design and applicable release decisions.

Trello

Requirements, acceptance criteria, work status and signoffs.

Git

Versioned source, approved asset provenance, design records and release evidence.

Vercel

Website deployment and serving the founder page and its versioned portraits.

Supabase

The active founder identity record queried by the website.

The publishing operation

1. 01Atlas: scope and evidence
2. 02Vex: branded design
3. 03Null: design review
4. 04Forge + Git: implementation
5. 05Preview and checks
6. 06Null: applicable release decision
7. 07Vercel: deployment
8. 08Production verification
9. 09Git: release evidence → scope and evidence

This diagram describes the review path for this work. Atlas review remains advisory; it adds no extra approval gate.

For the existing founder page, the website queries Supabase for identity data and selects the approved portrait from versioned website files. Current source and Founder One’s live HTML were checked; an exact deployed-commit audit and fresh rendered phone review remain separate checks.

The portrait correction is a useful example: the initial HTML retained old images; versioned approved portraits corrected the presentation while preserving the printed QR route. The lesson is practical: check the actual result, as well as the build.

The Founder One conversation

Which first customer and recurring task would you choose? What would you need to see to believe this could become a useful business—and something you would want to help build?

Sources: official provider descriptions reviewed October 15, 2026: RentAHuman, RobotFail, Virtuals ACP, Agentverse, Moltbook. These describe advertised capabilities, not independent performance audits. Pipeline evidence is recorded in the linked project DESIGN.md and founder-identity project records.