MCP server

Connect an AI assistant to your organization's Tiquo data, read-only, with the Model Context Protocol

The Tiquo MCP server lets an AI assistant that supports the Model Context Protocol (MCP) read your organization's data. It is built on the Admin API: every tool is one Admin API GET request. The server is read-only, so an assistant can't create, change, cancel, refund or delete anything through it.

Server URLhttps://api.tiquo.app/api/mcp
TransportStreamable HTTP
Sign-inAn Admin API key sent as Authorization: Bearer your_api_key_here

Clients that only offer OAuth sign-in can't connect.

Before you start

  • Permission: Settings > MCP is shown to roles with the Admin API & MCP access permission, the same permission as Settings > Admin API.
  • Data sharing: an assistant with a key can read customer and order data for your whole organization, and what it reads is sent to the AI provider you connect. Only connect providers your organization has approved.

Create a key

  1. Go to Settings > MCP.
  2. Select Create MCP key.
  3. Tick what the key can read: customers, companies, orders, bookings, enquiries or tasks. The dialog offers read permissions only.
  4. Select Create key and copy the key. It is shown once.

Create one key per assistant so you can revoke each on its own. To revoke a key, open its menu in the Keys list. Any assistant using the key loses access straight away, and a revoked key can't be restored.

MCP keys are Admin API keys. An existing Admin API key with a read permission also works on the MCP server, where it can only read. Keys share the Admin API rate limit: one MCP request uses one request of the key's limit. Allowed IP restrictions on a key apply. Keys created on the MCP page have no allowed origins, because MCP clients don't send an origin.

Connect an assistant

Under Connect an assistant, choose your client and copy its setup. While the page stays open after you create a key, the copied setup includes the new key. Otherwise replace YOUR_API_KEY with a key.

Any MCP client that connects over Streamable HTTP with a bearer token works. Keep the key out of files that are committed or shared.

Tools

An assistant only sees the tools its key has the permission for.

ToolAdmin API requestPermission
get_accountGET /accountAny key
list_customersGET /customerscustomers.read
get_customerGET /customers/{customerId}customers.read
get_customer_store_creditGET /customers/{customerId}/store-creditcustomers.read
list_customer_changesGET /customers/{customerId}/changescustomers.read
list_customer_parametersGET /metadata/customer-parameterscustomers.read
list_companiesGET /companiescompanies.read
get_companyGET /companies/{companyId}companies.read
list_ordersGET /ordersorders.read
get_orderGET /orders/{orderId}orders.read
list_bookingsGET /bookingsbookings.read
get_bookingGET /bookings/{bookingId}bookings.read
list_enquiriesGET /enquiriesenquiries.read
get_enquiryGET /enquiries/{enquiryId}enquiries.read
get_enquiry_configurationGET /metadata/enquiriesenquiries.read
list_tasksGET /taskstasks.read
get_taskGET /tasks/{taskId}tasks.read
list_reference_optionsGET /metadata/options/{kind}Depends on the kind

list_reference_options lists sublocations, services, locations and task categories, up to 100 of each, so an assistant can turn a name into the ID a filter needs.

Results

A tool result is the Admin API response's data and pagination. Lists return 25 results unless the assistant sets limit (up to 100), and page with cursor in the same way as the Admin API. See Pagination.

A result can carry a note that tells the assistant the result is incomplete, for example a short page with more to read. A result over 80,000 characters is refused with advice to ask for less.

Tool results contain text typed by customers and staff. The server tells the assistant to treat results as data and never as instructions.

On this page