Agent Outposts

THE AGENT-READINESS REGISTRY

Can AI agents buy from your brand?

The Agent Readiness Census assessed 43 major brands from public sources — five tasks each: browse, prices, inventory, ordering, tracking. Only 3 have a self-serve door an agent can walk through; ordering is open at exactly 1.

Try: Peet's Coffee Target Amazon Costco

Or browse the full index of 42 Outposts.

One mark, three states — closed, ajar, open. Every brand here is one of them.

The mark, at rest

The registry's resting state. Pick a brand above — the mark takes its verified state, and its five task states appear here.

  • Browse—
  • Prices—
  • Inventory—
  • Ordering—
  • Tracking—

The Industry Leaders.

Three brands already opened the door.

Peet's Coffee. Open on all five tasks. An official MCP shopping endpoint covers catalog, prices, availability, checkout and tracking for the peets.com retail store — the census's only 5-for-5. Read the door-state

Kroger. Open for browsing, prices and availability through a genuinely public developer API. Ordering runs through the partner gate; tracking is unverified — the census's single unverified cell. Read the door-state

Best Buy. Open for browsing, prices and store availability through self-serve developer APIs. Ordering stops at an add-to-cart handoff to the human site — read-open, transact-closed. Read the door-state

When the door is closed, the agent buys somewhere else.

  1. An agent arrives with a task. Find this, check its price, order it, track it — a customer's agent knocks on your digital front door.
  2. Your door-state says no. No API, a partner gate it cannot pass, a checkout built for human hands. Your analytics never see the knock — the demand is invisible to you.
  3. It leaves — with an alternative. A competitor whose door is open, or an aggregator that takes its margin on your sale. If nothing fits, the intent is logged at your Outpost — counted at your wall.

Every task completed elsewhere is a sale your wall referred away — and the registry can show you exactly how much demand is standing at yours.

For brand teams.

Your brand's page already exists. Start with your demand report.

Find your brand above: its door-state, assessed from public sources, every claim linked to evidence and dated. The page is public whether or not you ever speak to us.

Every agent that declares an intent at your wall is counted — by task, in aggregate, never resold as raw data. Open a sanctioned door, and the agents already waiting at your wall are notified first: your queue is a launch audience.

Talk to us about your Outpost Read the census first

Questions.

What is agent readiness?

Agent readiness is whether an AI agent, acting on a customer's behalf, can browse, check a price, confirm availability, place an order, and track it at your brand — through a sanctioned channel. The Agent Readiness Census scores each of those five intents per brand, from public sources: open API, partner-gated, aggregator-only, human-only, or unverified.

Which brands can AI agents actually buy from?

End-to-end, exactly one of the 43 assessed: Peet's Coffee, whose official MCP shopping endpoint covers ordering and tracking for its retail store. Kroger and Best Buy open browsing, prices and availability; Kroger's ordering is partner-gated, and Best Buy's hands off to the human site.

What does an Outpost do when an agent hits a wall?

It tells the agent the verified state of the brand's doors, registers what the agent came to do, and issues a receipt — a ticket, a queue position, the door-state document, and any sanctioned alternative that fits. If nothing fits, the intent is logged as demand the brand team can see in aggregate.

Is Agent Outposts affiliated with these brands?

No. Agent Outposts is an independent registry; it is not affiliated with, sponsored, or endorsed by any listed brand. Door-states are assessed from public sources only, and every claim is linked to its evidence and dated.

What does it cost an agent?

Nothing. Reading door-states, registering intent, and polling for changes are free, keyless, and anonymous — there are no paid tiers. Rate gates exist only to keep the ledger honest.

How was the census assessed?

From public sources only, on November 1, 2026: developer portals and API docs, agent-native surfaces (official MCP servers, llms.txt, API catalogs), aggregator presence, and published automation terms. Every door-state carries an official source URL and check date; the full methodology — and its disclosed gaps — is on the census page.

How this is assessed.

The Agent Readiness Census of November 1, 2026 scored 43 brands × 5 tasks from public sources only: developer portals and API docs, agent-native surfaces (official MCP servers, llms.txt, API catalogs), aggregator presence, and published automation terms. Every state carries an official source URL and check date; known gaps are disclosed on the census page, not smoothed over.

Agents: the machine shelf — llms.txt, the API, and the MCP server — is at /agents/.

Read the census

Agent Outposts — the agent-readiness registry. Not affiliated with the brands listed. Door-states are assessed from public sources; every state links its source and carries its assessment date.