Turbo Notify in Hermes

In Hermes Agent, Turbo Notify comes in as an MCP server, set up in config.yaml, and the virtual attendant skill is installed with a command. The attendant works in assisted mode: a person approves every answer, and nothing is answered while Hermes is closed.

What you need

A Turbo Notify account with a connected number (quickstart).
Hermes Agent installed.
For the skill: Python 3.9 or newer.

Connect

The MCP server, in ~/.hermes/config.yaml:

mcp_servers:
  turbo_notify:
    url: "https://mcp.turbonotify.com"
    auth: oauth

The skill:

hermes skills install turbo-notify/agent-skills/skills/virtual-attendant

After changing config.yaml, run /reload-mcp in an open session.

With your account: the auth: oauth above makes Hermes ask for the sign-in. Because signing in takes longer than Hermes waits inside a session, run this in another terminal:
```
hermes mcp login turbo_notify
```
Sign in to Turbo Notify and approve access.
With an API key: create the key in API Keys (opens in a new tab) and replace auth: oauth with a header:
config.yaml
```
mcp_servers:
  turbo_notify:
    url: "https://mcp.turbonotify.com"
    headers:
      Authorization: "Bearer YOUR_KEY"
```
Treat the key like a password.

Scopes and keys are in MCP server.

Test

Reload with /reload-mcp and ask Hermes: “List my WhatsApp numbers on Turbo Notify.” It calls list_numbers, which only reads. Nothing is sent.

For the skill, ask: “Set up the Turbo Notify virtual attendant.”

Update

The MCP server needs no update: the address is always the same. For the skill, run the install command again.

Limits

Assisted mode only, with you present. Every answer needs approval.
The tools follow the API limits: 600 requests per minute per key or sign-in, plus the plan’s quota and sending pace. See MCP server limits.

What is not verified

The MCP server in config.yaml and signing in with hermes mcp login. The format and the commands follow Hermes’s documentation, and we did not run the full flow for this page.
The skill inside Hermes, and updating by installing over it. We have not run the attendant in this agent.