Hermes runs on its own for hours. Don’t leave a token on its disk.
Hermes is a full MCP client, so it talks to AgentValet the way it talks to any MCP server. Connect it over OAuth and it gets an identity, not a credential: its config holds a URL and nothing secret.
Paste one snippet, approve in the browser
The dashboard builds the snippet from your owner id. There’s no token, bearer or API key in it.
In the dashboard, open Agents, click Connect Claude, ChatGPT, Copilot… and pick the Hermes tab.
Paste the block into
~/.hermes/config.yaml.~/.hermes/config.yaml mcp_servers: agentvalet: url: https://api.agentvalet.ai/mcp/v1/<owner-id> auth: oauth timeout: 60Start Hermes. On first use it opens your browser; sign in and click Approve. AgentValet creates a new agent for it, tagged as Hermes in the audit log.
Grant that agent its platforms in the dashboard.
Add it from the command line instead
The Hermes tab also gives you the equivalent command, built from the same owner id. It still connects over OAuth.
Run it, then start Hermes and approve in the browser.
Terminal hermes mcp add agentvalet --url https://api.agentvalet.ai/mcp/v1/<owner-id>
What’s governed here
Nothing is allowed until you grant it. This is a typical first grant for Hermes; you decide every row, and you can change it without touching the agent.
| What the agent tries | Platform | Decision |
|---|---|---|
| Read your calendar | Google Calendar | Allowed within grant |
| Create a task | Todoist | Allowed within grant |
| Send an email | Gmail | Asks you |
| Anything you haven’t granted | Any | Not granted |
Hermes guide in the docsThe longer walkthrough, kept with the product docs.
Put Hermes behind AgentValet.
Register the agent, grant only what it needs, and approve the risky calls from your phone. Free to start, no credit card.