Connect a client

Sign in as yourself

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

Add --scope user for every project, or --scope project to share it through .mcp.json. Start claude, run /mcp, select telflo, choose Authenticate, and finish in the browser. On a machine without a browser, newer releases offer claude mcp login telflo, which prints the sign-in URL and takes the redirect URL back at the prompt.

Use a service-account token

{
  "mcpServers": {
    "telflo": {
      "type": "http",
      "url": "https://mcp.telflo.com/mcp",
      "headers": {
        "Authorization": "Bearer ${TELFLO_API_TOKEN}"
      }
    }
  }
}

Claude Code expands ${TELFLO_API_TOKEN} from each person's environment, so the file can be committed; an unset variable is sent as literal text and fails with a 401. Passing the token to claude mcp add --header instead writes its value to ~/.claude.json.

See Claude Code's MCP documentation.

claude.ai and Claude Desktop add remote servers as custom connectors on your Claude account.

Sign in as yourself

1. Open Customize > Connectors, at https://claude.ai/customize/connectors or from Customize in Claude Desktop's sidebar, and select + Add, then Add custom connector.
2. Enter a name, such as Telflo, and the URL https://mcp.telflo.com/mcp, and select Continue twice.
3. Under Authentication, choose Sign in now; under OAuth client, choose Register automatically. Select Add and finish in the browser.
4. In a chat, turn the connector on from the + button, under Connectors.

On a Team or Enterprise plan, an Owner adds the connector under Organization settings > Connectors, and each member selects Connect on it under Customize > Connectors to sign in as themselves.

Use a service-account token

Claude sends a fixed header only through the Request headers section of Add custom connector, a beta not every organization has. Where it is available, set Authentication to No sign-in and add a required authorization header with the full value Bearer tf_.... Everyone in your Claude organization who uses the connector then acts as that one service account, so use a read-only token unless all of them should be able to push and deploy.

See Anthropic's guide to custom connectors.

Sign in as yourself

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

Use ~/.cursor/mcp.json for every project or .cursor/mcp.json for one. With no headers, Cursor registers itself and opens the browser. Cursor desktop returns to http://localhost:8787, but when that port is busy, and in some older releases, it asks for a cursor:// link, which the Authorize page refuses with The application supplied an unsafe redirect URL.

Use a service-account token

{
  "mcpServers": {
    "telflo": {
      "url": "https://mcp.telflo.com/mcp",
      "headers": {
        "Authorization": "Bearer ${env:TELFLO_API_TOKEN}"
      }
    }
  }
}

Cursor substitutes the variable from the environment it starts from.

See Cursor's MCP documentation.

MCP servers work in GitHub Copilot's agent mode in VS Code 1.102 or later. On Copilot Business or Enterprise, an administrator first turns on the MCP servers in Copilot policy.

Sign in as yourself

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

For your user rather than the workspace, run MCP: Open User Configuration from the Command Palette. VS Code opens the browser the first time it connects.

Use a service-account token

{
  "inputs": [
    {
      "type": "promptString",
      "id": "telflo-token",
      "description": "Telflo API token (tf_...)",
      "password": true
    }
  ],
  "servers": {
    "telflo": {
      "type": "http",
      "url": "https://mcp.telflo.com/mcp",
      "headers": {
        "Authorization": "Bearer ${input:telflo-token}"
      }
    }
  }
}

VS Code asks for the token once and stores it securely. Use the input rather than ${env:TELFLO_API_TOKEN}: an open VS Code bug sends an ${env:...} header reference as literal text, which fails with a 401.

See VS Code's MCP configuration reference.

Sign in as yourself

codex mcp add telflo --url https://mcp.telflo.com/mcp

Codex detects Telflo's sign-in and opens the browser; if it asks you to, run codex mcp login telflo, which also signs in a server added to ~/.codex/config.toml by hand.

Use a service-account token

export TELFLO_API_TOKEN=tf_...
codex mcp add telflo --url https://mcp.telflo.com/mcp --bearer-token-env-var TELFLO_API_TOKEN

or in the configuration file:

[mcp_servers.telflo]
url = "https://mcp.telflo.com/mcp"
bearer_token_env_var = "TELFLO_API_TOKEN"

The variable holds the token alone; Codex adds Bearer and stores only the variable's name.

See Codex's MCP documentation.

Sign in as yourself

gemini mcp add --scope user --transport http telflo https://mcp.telflo.com/mcp

--scope user writes to ~/.gemini/settings.json instead of the project's .gemini/settings.json. Gemini CLI opens the browser when it first connects, and /mcp auth telflo signs in again. The browser must run on the same machine.

Use a service-account token

{
  "mcpServers": {
    "telflo": {
      "url": "https://mcp.telflo.com/mcp",
      "type": "http",
      "headers": {
        "Authorization": "Bearer $TELFLO_API_TOKEN"
      }
    }
  }
}

Gemini CLI replaces $TELFLO_API_TOKEN from the environment when it loads its settings; an unset variable is sent as literal text.

See Gemini CLI's MCP documentation.

Windsurf is now Devin Desktop. From version 3.9.19, its agent reads MCP servers from the Devin CLI's configuration: ~/.config/devin/mcp_config.json for your user, or .devin/mcp_config.json in a project. Earlier releases, with the Cascade agent, read ~/.codeium/windsurf/mcp_config.json and use serverUrl in place of url.

Sign in as yourself

devin mcp add telflo https://mcp.telflo.com/mcp
devin mcp login telflo

Without devin mcp login, Devin asks you to sign in the first time it uses the server.

Use a service-account token

devin mcp add has no header option, so write the header into the project's local file, which Git ignores:

{
  "mcpServers": {
    "telflo": {
      "url": "https://mcp.telflo.com/mcp",
      "transport": "http",
      "headers": {
        "Authorization": "Bearer tf_..."
      }
    }
  }
}

See Devin's MCP configuration documentation.

ChatGPT connects only through the browser sign-in and cannot send a token. Your plan and workspace settings decide whether you can add a custom MCP server and whether its write tools can run.

1. Go to https://chatgpt.com/plugins, select the plus button, then Add custom MCP server.
2. Enter a name, set Server URL under Connection to https://mcp.telflo.com/mcp, and choose OAuth under Authentication.
3. Select I understand and want to continue, then Create as a plugin, and complete the Telflo sign-in. The Authorize page shows chatgpt.com.
4. Install the plugin from your personal plugins, and mention it with @ in a prompt.

OpenAI also documents an older developer-mode route, which an admin enables first on Business, Enterprise, and Edu workspaces.

See OpenAI's guide to adding a custom MCP server.

The MCP Inspector connects to a server and calls its tools by hand. Version 2 needs Node.js 22.19 or newer; tutorials written for version 1 show different flags.

Sign in as yourself

npx @modelcontextprotocol/inspector --server-url https://mcp.telflo.com/mcp --transport http

Open the URL the Inspector prints and select Connect. It registers itself and opens Telflo's sign-in.

Use a service-account token

export TELFLO_API_TOKEN=tf_...
npx @modelcontextprotocol/inspector --server-url https://mcp.telflo.com/mcp --transport http --header "Authorization: Bearer $TELFLO_API_TOKEN"

To list the tools without the web interface:

npx @modelcontextprotocol/inspector --cli https://mcp.telflo.com/mcp --transport http --header "Authorization: Bearer $TELFLO_API_TOKEN" --method tools/list

See the Inspector's documentation.