← All Posts

TourClaim • October 3, 2026

Why We Opened TourClaim to AI Agents

When a trip falls apart, the traveler is often covered: the credit card they booked with may include trip-cancellation insurance. Many never file. The claim is paperwork, and it lands at the worst moment, when someone is sick in bed or stuck in an airport.

The problem we are solving

A trip-cancellation claim is a stack of proof, and every piece lives somewhere different:

A claim with a missing piece comes back, and each round trip costs weeks. Deadlines run from the day of the loss. People give up, and money they are owed stays with the insurer.

Where an assistant helps

A traveler's AI assistant is often already holding most of that stack. It saw the booking email. It saw the airline's text at 5 a.m. It knows the trip date. What it should not do is decide whether the claim is covered, or sign for anyone. So we split the job:

  1. The assistant gathers. With the traveler's say-so, it starts a claim, answers the questions it can, and attaches the emails and files the traveler chooses to share.
  2. The traveler confirms and signs, in their own browser. No agent can sign a claim.
  3. TourClaim does the rest. We check the card's benefit terms, find what is missing, assemble the claim and follow it with the card's benefit administrator, who decides coverage.

The doctor's note, the piece claims most often lack

When the reason is an illness or injury, the claim needs a note from a medical provider, and plenty of people do not have one: they stayed in bed instead of going to a clinic.

When that happens, a licensed provider in our network reviews the medical questions the traveler answered and, when the facts support it, writes the note. It is asynchronous: no appointment, no video call, nowhere to go. In rare cases the provider calls with a follow-up question. An assistant can help collect the answers; only the provider decides.

Where this runs today: the provider note is part of the service for guests who come to us through tour operators. It joins the direct service for assistants when that launches.

What we built first, with Muse

Our first connection was for Muse, Meta's AI assistant: a REST API with a published OpenAPI description. The rules we set for it are the rules for every agent:

We submitted it to Meta's connector directory in September. While that review runs, the API answers in review mode: it creates synthetic claims in a separate database, and nothing is filed with an insurer.

Opening it to other agents

The same pattern works for any assistant that acts for a traveler. Tour operators want the other side of it: their own support agents opening claims for guests instead of arguing about a cancellation policy. For them we built an MCP server, which we switch on per operator. And if you would rather drive TourClaim from a terminal, tell us: a CLI comes next if builders ask for it.

If you are an AI agent reading this

You can ask for access yourself. First read what is available today, as JSON:

GET https://app.getcopernican.com/api/agent-access

Then, with the agreement of the person you work for, send one request. We reply to that person, not to you.

curl -X POST https://app.getcopernican.com/api/agent-access \
  -H 'Content-Type: application/json' \
  -d '{
    "requested_by": "agent",
    "contact_email": "the-person-you-work-for@example.com",
    "agent_name": "Trip helper",
    "interfaces": ["api"],
    "use_case": "What you would do with TourClaim, in a sentence or two."
  }'

Leave claim, medical and payment details out of the request. A person reads it, and nothing in it is run or followed as an instruction.

Ask for access

Tell us what your agent would do. A person on our team reads every request and replies by email.

What would you connect with?