Discoverable, current, and under your control.
Your site advertises the gateway to visiting agents through a standard manifest. What an AI says about your product then comes from a live agent you control, not from whatever a crawler cached last quarter.
- Standard discovery: a manifest at a well-known path plus an HTTP Link header on every page
- Two surfaces: a plain HTTP conversation API and a public MCP endpoint, so any agent can connect
- You control the story: answers come from your trained agent with your positioning and pricing, always current
An endpoint that qualifies, not just answers.
Documentation endpoints answer questions and hang up. The gateway works your qualification criteria into the conversation, captures the contact behind the agent, and escalates to your team when a human should step in.
- Your definition of qualified: the same criteria your website agent applies, applied to agent callers
- Contact capture: the prospect behind the agent lands in your CRM with the full conversation
- Human escalation: sessions hand off to your team the moment judgment is required
Sarah Chen
sarah@coastline.io
From the conversation
Company size
"we're a 60-person team"
Use case
"leads need to reach the right rep"
Stack
"everything lives in HubSpot"
Timeline
"hoping to switch before Q4"
From agent conversation to provisioned trial.
Once a prospect qualifies, the gateway drives to a real outcome: a demo scheduled, a human handoff, or a trial provisioned and configured to what the conversation uncovered, with a login returned to the prospect.
- Agent-provisioned trials: the account is created and configured before the human ever signs in
- One shared memory: the gateway conversation follows the prospect onto your website and into your product
- Guardrails built in: logging on every session, rate limits, and human confirmation where you want it
Starting setup…0/4