Claude Desktop quickstart

Claude Desktop supports remote MCP servers with browser-based OAuth consent. AgentValet is a native OAuth 2.1 authorization server, so the app connects directly: AgentValet provisions a new agent for the connection, and Claude holds a short-lived access token from AgentValet’s OAuth server rather than a signing key. There’s nothing to paste and nothing sitting on disk.

Requirements

Claude Desktop installed
An AgentValet account (free) at app.agentvalet.ai

Set up

Sign in at app.agentvalet.ai and copy your owner ID from Settings → Account.
In Claude Desktop, open Settings → Connectors and add a remote MCP server with this URL:
```
https://api.agentvalet.ai/mcp/v1/<owner-id>
```
Approve the OAuth consent in your browser. AgentValet provisions a new, deny-by-default agent for the connection.
Connect and scope platforms in the dashboard . The app can now call them through the governed proxy.

OAuth connections aren’t plan-gated, this works on every tier including Free.

Prefer a config file?

You can also add an agentvalet entry to claude_desktop_config.json instead of using the Connectors UI:


{
  "mcpServers": {
    "agentvalet": {
      "url": "https://api.agentvalet.ai/mcp/v1/<owner-id>"
    }
  }
}

Alternative: the MCPB extension

If you’d rather not connect over OAuth, the MCPB extension is the other supported path for Claude Desktop, and it gives the agent its own signing key: download agentvalet-latest.mcpb, drag it into Claude Desktop’s extensions window, then register a Claude Desktop agent from the dashboard’s Agents page and paste the agent ID, owner ID, proxy URL and private key it shows into the extension’s install dialog. See Register an agent (MCPB). See Install paths compared for the full comparison.

Confirm it loaded

Open Settings → Connectors (or the extensions panel, if you used MCPB) and confirm agentvalet shows as connected.

Make your first governed call

Ask Claude to call list_platforms. It should return the platforms approved for this agent, an empty list at first, and that’s fine. What matters is that the call succeeds: the connection authenticates correctly, the proxy is reachable, and the agent record is active.

From there, grant scopes for a platform you’ve connected in the dashboard and try a real call.