A landing page agents can actually use

Most product landing pages are for humans: a headline, a screenshot, a form. An agent hitting that page cannot click the demo video or interpret a hero image. If the product is a mailbox for software, the page has to work as an interface for software: a stable URL, a machine-readable way to provision, and an API that does not require a marketing site crawl.

Pattern Automation’s Agent Inbox is API-first. Humans still get a site they can read. Agents get SDKs, MCP, and a create call. Neuro OS roles can provision an inbox without pretending to be a user with a mouse.

What humans need versus what agents need

Humans need to understand the job: each role gets an address people can reply to, mail can forward into Neuro OS, outbound defaults to Ask. They need pricing context, examples, and a way to talk to us.

Agents need:

• A documented base URL and auth scheme.
• A create-inbox operation that returns an id.
• Tools to list threads, read, search, and draft.
• Webhooks with verification.
• No requirement to solve a carousel.

If your only onboarding is “sign up and click around,” agents will fail or they will invent a browser Rube Goldberg. Offer MCP and SDK as first-class paths, not as a footer link.

Provisioning is a skill, not a screenshot

A Neuro OS skill can say: create an inbox for this role, store the id, attach MCP, set send to Ask. The code is small:

from agentinbox import AgentInbox
client = AgentInbox()
inbox = client.inboxes.create(username="hello", domain="agentinbox.space")

That snippet belongs on the landing page because it is the product. A dashboard screenshot of an empty inbox is not. Agents copy the call; humans see that the call is the onboarding.

Username and domain should be policy-driven. The skill should not let a model pick ceo on a shared domain without an admin gate. Creating mailboxes is closer to creating infrastructure than to sending a chat message. Ask applies.

Machine-readable, not hidden

Publish an OpenAPI document or MCP tool list at a stable path. Version it. When you change a field name, agents break silently and then send the wrong thing — except they cannot send, because Ask is on, and then a human stares at a broken draft. That is better than a wrong send, but it is still downtime.

The landing page can point at introducing Agent Inbox for the narrative and keep the hero focused on the create call. Internal docs should show threading — email threading for AI agents — because agents that cannot stay on a thread are not using mail; they are generating new subjects.

Forwarding and Ask are part of the page

If you explain only “agents get email,” teams will auto-send. State the default: outbound is Ask; CRM writes are Ask; humans receive forwarded threads when the role is stuck. That copy is for people, but it also belongs in the skill the agent loads so the model does not treat send as success.

A landing page agents can use is still a page for buyers. The difference is that the primary action is an API, and the primary metaphor is a role mailbox, not a human webmail clone. Build the site for both audiences. Ship the mailbox as code.

Agent Inbox gives each role a mailbox people can reply to, with forwarding into Neuro OS when a human must see the thread. Outbound mail defaults to Ask. Run the role on Neuro OS. To scope the first inbox, get started.

Explore Neuro OS →