Skip to main content
We provide an MCP (Model Context Protocol) server with prompts, resources and tools that you can use for building AI-applications. The server uses streamable HTTP transport providing real-time communication with Server-Sent Events (SSE) streaming support.

Getting Started

Our MCP server is located at https://mcp.tic.io. It accepts your TIC API key via either of two headers:
  • x-api-key: <your_api_key> - same scheme as our REST API
  • Authorization: Bearer <your_api_key> - for clients (such as the Anthropic Messages API connector) that only forward an OAuth-style bearer token
Either form authenticates the same way against the same key. Pick whichever your MCP client supports.

Example of using OpenAI Responses API

OpenAI Responses API is the most advanced interface for generating model responses. Full documentation for the API is available here. Endpoint: https://api.openai.com/v1/responses Note you need to include Bearer authentication for the OpenAI to work. An example of the request body follows below that is using flagship gpt-5 model.

Example of using Anthropic Messages API

Anthropic’s Messages API can call remote MCP servers directly through its MCP connector. Full documentation is available here. Endpoint: https://api.anthropic.com/v1/messages You need three headers on the request:
  • x-api-key: <your_anthropic_api_key> - your Anthropic API key
  • anthropic-version: 2023-06-01
  • anthropic-beta: mcp-client-2025-11-20
The authorization_token field on the MCP server entry is forwarded by Anthropic to our server as Authorization: Bearer <token> - pass your TIC API key there. An example request body:
To restrict which TIC tools the model can see, use the mcp_toolset allowlist pattern - disable everything by default, then enable specific tools:

Example of LibreChat configuration

If you wish to use our MCP in LibreChat you simply add a block inside mcpServers section in yaml configuration file:
librechat.yaml

Example of prompt and response

This shows an example of a user input prompt and the response. The request is using gpt-4.1. Clean page

Prompts

Prompts are user-invokable workflow templates that combine multiple tools into a single investigation flow with explicit framing and a named deliverable. Pick one from your MCP client’s prompt menu to kick off an end-to-end analysis. Most prompts take registrationNumber (Swedish organisationsnummer, e.g. 556519-9493) and resolve identity to the internal companyId as their first step. build_prospect_list is the exception - it starts from a free-text brief rather than a known company.

Resources

Resources are static reference data the model reads once into context - no tool call required. They decode the codes that appear in tool responses.

Tools

Functions the model invokes to retrieve company, property, or vehicle data. Tools are grouped below by purpose.

Company search & identity

Company financials & risk

Company enforcement & compliance

Company structure & history

Vehicles

Property

Prospecting & registers

search_companies_for_leads

The broadest tool on the server. It combines free-text matching against company name, business purpose (verksamhet) and SNI industry names with structured firmographic filters, and returns a paged lead list. Size filters are bucketed: pass a number and it maps to the SCB band containing it. Sorting by turnover or employee count is not available, because the SCB size-class field is not sortable - filter on it instead.

search_lei

We will continuously add new tools - check back here for new additions.