DayMarket: the boarding station prototype.

The problem

AI agents are about to start placing orders, booking appointments, and requesting quotes. They are prompts that go retrieve the software they need and perform tasks. But a business like a donut shop or a plumbing company is invisible to them: no structured menu, no ordering API, no machine-readable anything. The agent cannot see them, so the agent cannot buy from them.

Dashboards do not solve this. The agent never looks at a dashboard.

What we built

Readable Station: the boarding station for the agent economy. Not a directory. A directory lists; listings rot and nobody transacts on them. A station is where an agent arrives and actually does business: discovers the capability, invokes it, pays, gets confirmation.

Three layers:

1. Legibility. The business is machine-readable: services, prices, hours, policies. Structured data an agent can consume.
2. The gate. An agent-invokable endpoint for the business's key workflow. The agent invokes it, payment moves, confirmation comes back. The owner never touches anything. They get the order.
3. Verification. Every transaction through the station writes a verification record. When agents act invisibly, audit is the moat. This is the seed volume for Readable Evals.

The prototype: DayMarket

DayMarket proved the loop before we named it. A buyer places a catering order. The business confirms and delivers. The 15% fulfillment fee moves automatically. The business sees a balance and cashes out on demand. Nobody on the business side touches a dashboard.

The Station build made it agent-invokable:

• /.well-known/agent.json: the station manifest. Every capability listed with JSON schemas. The departure board.
• POST /api/agent/order: the first gate. JSON in, order confirmation out, routed through the existing order flow. Idempotent, authenticated, validated.
• llms.txt and JSON-LD on business profiles: the legibility layer.
• Verification records on every agent order: the audit trail and the Evals seed.

Local validation: 11 of 11 checks passed, covering the happy path, authentication, input validation, unknown-entity handling, and idempotency.

The wedge

This is the same funnel Readable already runs:

1. The $150 assessment finds the costliest workflow.
2. The Sprint maps it and prices the leak.
3. The Station turns that workflow into a gate any agent can invoke.

One workflow was always the promise. Now the workflow becomes infrastructure, not just a fix.

What is next

First vertical: catering and ordering, proven. Next: booking (chairs, tables, bays), then quotes (trades). Every gate adds verification volume, and every verification record makes the Evals enterprise case stronger.

Start with the $150 assessment