What MCP is and how to connect Claude to your business data

Until recently, connecting an AI assistant to a company's systems meant building a custom integration for every tool. The Model Context Protocol (MCP) changes that: it defines a standard way for a model to discover which tools are available, which parameters they accept and what they return.

What MCP solves

A language model knows nothing about your clients, your tasks or your documents. MCP gives it a catalogue of actions (search, read, create, update) and the model itself decides when to use them. Unlike copying and pasting information into a chat, the assistant works with up-to-date data and its actions are logged.

MCP doesn't give AI access to everything: it gives it access to the tools you expose, with the permissions of the person it represents.

How Unelia connects

Unelia publishes an MCP server that exposes the tools in your workspace. If your assistant supports OAuth connectors, like Claude, just add the address https://public.app.unelia.com/v1/mcp-oauth and sign in with your account. If you would rather connect it with an API key, add this server configuration:

{
  "mcpServers": {
    "unelia": {
      "type": "http",
      "url": "https://public.app.unelia.com/v1/mcp",
      "headers": {
        "x-api-key": "prefix.secret"
      }
    }
  }
}

When you restart Claude, you'll see Unelia's tools available. From then on you can ask things like "which proposals have gone a week without a reply?" or "create a task to call Talleres Puig on Friday".

Permissions and logging

• The API key belongs to a person: the AI never has more permissions than they do.
• Every call is recorded in the activity log as an action delegated to an AI.
• You can revoke the API key at any time from Settings.

Next steps

Start with read-only tools and, once the team feels comfortable, enable task and note creation. The developer documentation includes the full list of available tools.