Tero MCP Server

Tero exposes a Model Context Protocol (MCP) server, letting external AI tools — such as Claude Code, Claude Desktop, or Cursor — interact directly with your Tero agents.

Server URL

The MCP server is available at your Tero instance URL with /mcp appended, for example: https://your-tero-instance.example.com/mcp.

Tero's MCP server uses OAuth via OpenID Connect. When you connect a client for the first time, the client will open a browser window for you to log in with your Tero credentials. After authorizing, the client stores the session and subsequent requests are authenticated automatically.

Client Configuration

Claude Code

You can add the Tero MCP server either via the CLI or by editing the configuration file directly.

Via CLI:

claude mcp add --transport http tero https://your-tero-instance.example.com/mcp

Via JSON (~/.claude.json for user scope, or .mcp.json in project root for project scope):

{
  "mcpServers": {
    "tero": {
      "type": "http",
      "url": "https://your-tero-instance.example.com/mcp"
    }
  }
}

After saving the JSON file, open a Claude Code session and run the /mcp command to connect and authenticate:

Type /mcp in the Claude Code prompt and press Enter.
Claude Code will list your configured MCP servers and show their connection status.
If the Tero server shows as Not connected, select it and choose Authenticate to start the OAuth login flow.
A browser window will open — log in with your Tero credentials and authorize the connection.
Run /mcp again to confirm the server status shows as Connected.

Claude Desktop

Open Claude Desktop and go to Settings → Developer → Edit Config.
Add the Tero server under mcpServers:

{
  "mcpServers": {
    "tero": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://your-tero-instance.example.com/mcp"
      ]
    }
  }
}

Save the file and restart Claude Desktop. On first use, it will open a browser window to complete the OAuth login.

Version requirement

Remote HTTP MCP servers with OAuth require a recent version of Claude Desktop. Make sure you are running the latest version.

Cursor

Open Cursor Settings → MCP and click New MCP Server.
Add the following under mcpServers:

{
  "mcpServers": {
    "tero": {
      "type": "http",
      "url": "https://your-tero-instance.example.com/mcp"
    }
  }
}

Open Cursor Settings → MCP and click Connect next to the Tero server.
A browser window will open to complete the OAuth login.

Refer to the Cursor MCP documentation for full details.

VS Code

Create or edit .vscode/mcp.json in your workspace root and add the Tero server:

{
  "servers": {
    "tero": {
      "type": "http",
      "url": "https://your-tero-instance.example.com/mcp"
    }
  }
}

Reload the VS Code window after saving. On first use, Copilot will prompt you to complete the OAuth login in your browser.

Version requirement

MCP support in VS Code requires version 1.99 or later with GitHub Copilot enabled.

Usage

Your AI client needs to be explicitly told to use Tero agents. For example:

"Use Tero to ask the support agent: what is the refund policy?"

"Ask the Tero QA agent to write test cases for this feature."

This tells the client to invoke the MCP tools and route the request through Tero.

Available Tools

Once connected, your AI client can use the following tools:

`list-agents`

List the Tero agents accessible to your account, optionally filtered by a search term.

Parameter	Type	Required	Description
`text`	string	No	Filter agents by name or description

Returns a list of agents, each with:

id — numeric agent identifier
name — agent display name
description — agent description

`start-conversation`

Create a new conversation thread with a specific agent. If an empty thread already exists for that agent, it is returned instead of creating a duplicate.

Parameter	Type	Required	Description
`agent_id`	integer	Yes	The ID of the agent to talk to (from `list-agents`)

Returns:

thread_id — identifier for the new thread, used in send-message
thread_name — display name of the thread (may be null until the first message is sent)

`send-message`

Send a message to an agent in an existing thread and receive the complete response. This is a synchronous call — it waits for the full agent response before returning.

Parameter	Type	Required	Description
`thread_id`	integer	Yes	The conversation thread ID (from `start-conversation`)
`message`	string	Yes	The message text to send

Returns:

response — the agent's complete text response
thread_id — the thread the response belongs to

Typical flow

Use list-agents to find an agent → start-conversation to open a thread → send-message one or more times to exchange messages.

`add-time-saving`

Record the estimated time saved while using an external AI agent. This is the MCP equivalent of the Register other agents button in the AI Console.

Parameter	Type	Required	Description
`agent_name`	string	Yes	Name of the external AI agent (e.g. `"Cursor"`)
`minutes_saved`	integer	Yes	Estimated number of minutes saved

This tool is typically called automatically by the AI client after each productive interaction, using a rule or instruction you configure.