Skip to main content

Use Terros with AI

Connect an AI client to Terros through the hosted Model Context Protocol (MCP) server. The server uses Streamable HTTP at:

https://mcp.terros.com/mcp

Sign in with your normal Terros login when your client opens the authorization flow. You do not need an API key, and compatible clients register themselves automatically.

In the Terros web app, Settings → Connect your AI shows this address with copy buttons and install links for each client below.

Access and permissions​

The connection is read-only by default. It can find and run operations that read Terros data, but it does not expose operations that change data.

To allow changes, connect to:

https://mcp.terros.com/mcp?writes=enabled

With writes enabled, tools act as you and can change data using your Terros permissions. Review tool calls carefully. Before an operation that deletes data runs, the server returns a preview and the AI client must confirm it.

Whether the connection is read-only or write-enabled, you can see and do only what your Terros role allows. The AI cannot get around your permissions or use operations your account can't use in the Terros app.

Customer and staff tools​

Every customer account gets the operations of the public Terros API, the same ones documented on this site. Some internal operations may be available to Terros staff accounts. Terros checks staff status on every request, so these never appear for a customer account.

How the tools work​

The server presents three tools instead of loading hundreds of individual operations into your AI client's context:

  • terros_search_tools finds the operation needed for a task.
  • terros_describe_tool returns that operation's inputs.
  • terros_call runs the operation as the signed-in user.

Operations are grouped into these toolsets: account, activity, area, attachment, calendar, company, connect, contact, custom-field, disposition, evaluation, export, external-link, game, import, location, permission, program, report, role, search, task, team, user and workflow.

You can change what a connection offers by adding options to the address:

  • ?toolsets=account,task limits the connection to those toolsets.
  • ?tools=endpoints lists one tool per operation instead of the three tools above. This uses much more of your AI client's context.

Combine options with &, for example https://mcp.terros.com/mcp?toolsets=account,task&writes=enabled.

Connect a client​

Not yet tested

None of these setup paths has been tested with Terros yet. They come from each vendor's current documentation. Product availability and menu names can depend on your plan and organization settings.

Claude Code (not yet tested)​

Run:

claude mcp add --transport http terros https://mcp.terros.com/mcp

Then run /mcp in Claude Code and follow the prompts to sign in. See Claude Code's MCP guide.

Claude desktop and web (not yet tested)​

In Claude, open Settings → Connectors, choose Add custom connector, enter https://mcp.terros.com/mcp, and select Add. In a chat, open Search and tools, select the connector, and choose Connect to sign in. Team and Enterprise organizations may require an owner to add the connector first. See Anthropic's custom connector guide.

ChatGPT (not yet tested)​

ChatGPT custom MCP apps require an eligible workspace and developer mode. Enable developer mode, then open Settings → Apps → Create, provide https://mcp.terros.com/mcp as the MCP endpoint, select OAuth authentication, and scan the tools. Complete the sign-in prompt, then create the app. See OpenAI's developer mode and MCP apps guide.

Cursor (not yet tested)​

Add a remote server to .cursor/mcp.json for one project or ~/.cursor/mcp.json for all projects:

{
  "mcpServers": {
    "terros": {
      "url": "https://mcp.terros.com/mcp"
    }
  }
}

Open Cursor's MCP settings and complete the OAuth sign-in when prompted. See Cursor's MCP guide.

VS Code (not yet tested)​

In VS Code, open the Command Palette, run MCP: Add Server, choose HTTP, enter https://mcp.terros.com/mcp, and name the server terros. Or add it to .vscode/mcp.json for one workspace:

{
  "servers": {
    "terros": {
      "type": "http",
      "url": "https://mcp.terros.com/mcp"
    }
  }
}

Start the server and sign in with your Terros account when VS Code asks. See VS Code's MCP guide.

Moving from terros mcp​

If your AI client runs the Terros CLI's local terros mcp server, replace that entry with the hosted address above. The hosted server needs no CLI install or developer mode and signs you in with OAuth.