Connect your AI client to Vampauth MCP

Updated October 27, 2026 · Vampauth Team · Client recipes shared with the dashboard.

Manage your licensing workspace from the tool where you code. Native Model Context Protocol (MCP) lets a compatible AI client use real Vampauth actions: inspect your projects, issue keys, reset hardware bindings and publish loaders when authorized.

Start with a read-only connection, confirm it with a tool call, then enable the management permissions your workflow needs. This guide uses the same client recipes as the dashboard.

1. Prepare your workspace

You need a Vampauth account and a compatible remote MCP client. If you are new, follow the first-project guide. For an existing account, open Dashboard → MCP to choose a client, view the endpoint and inspect your connections.

Use the endpoint shown in your dashboard. These recipes use this site's configured endpoint; if you connect to a different Vampauth environment, copy its endpoint from Dashboard → MCP and replace the address in the recipe.

https://vampauth.com/mcp

The endpoint is an authenticated protocol service, not a page to open for instructions. Browser OAuth does not require an API token in the client configuration. Keep existing MCP servers when merging the recipe into your client settings.

2. Add Vampauth to your AI client

Choose the recipe for your application. These examples use Compact mode: five tools discover the registered actions on demand. Classic mode exposes actions as individual typed tools; select it in the dashboard and reconnect if your workflow prefers it.

claude mcp add --transport http vampauth 'https://vampauth.com/mcp'

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

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

For another compatible application, add the endpoint as a remote Streamable HTTP MCP server with browser OAuth. Follow that client's configuration format; do not assume a normal consumer chat session supports remote MCP.

3. Sign in and verify a read-only request

Complete browser sign-in and review the requested permissions and project access. The default setup starts read-only. If the application needs a reload or reconnect to discover tools, complete that before testing.

Use the Vampauth vamp_context tool to report my account quotas and accessible projects. Use the tool result.

A successful response should come from a real tool call and identify the accessible account/projects or their actual limits. An authorization screen alone is not proof the client executed a tool. Inspecting key state must use management reads, not the game-client validation endpoint that can activate or bind a key.

4. Put MCP to work on your key system

The following are example requests, not recorded successful operations. Choose the intended project and authorize the required write scopes before asking a client to change it.

• “Find my project and show its current license capacity and publishing allowance.”
• “Create one 24-hour test key for this project, starting on first successful use.”
• “Reset this key's HWID binding without changing its expiry.”
• “Publish this Luau source as a loader with the runtime checks I selected.”
• “Show checkpoint completions, key usage and loader delivery counters, with their time-window limits.”

The client must discover and describe the real action, resolve canonical resource IDs, and read the current revision where required. Writes need a fresh operation ID per intended change. After an uncertain timeout, retry the same operation ID and identical inputs; do not blindly repeat the change with a new ID.

Publishing still requires source content, supported input sizes and available allowance. A remote server cannot read a file path on your computer. The separate local-file companion is not assumed installed or publicly available through npm.

5. Keep the connection scoped to your work

Connect a compatible AI client through Model Context Protocol (MCP) to manage projects, keys, loaders and providers. Connection scopes, current project permissions and resource allowances still apply; read-only actions do not activate keys. Consent can reduce a client's requested permissions, not invent permissions the client never requested. If a write is denied, inspect its required scopes and project access, then reconnect with the appropriate request if needed.

Provider credentials and webhook URLs are write-only in the applicable management contracts. Original source is not retained as a separate object after publishing. Read access, write access and file transfer are different capabilities; never put management credentials into a distributed game script.

To revoke access, open Dashboard → MCP → Your connections, choose Disconnect for the application and confirm. Revocation stops that connection's authorization; it does not undo earlier management changes.

If the connection or an action fails

• No tools: check the endpoint, remote HTTP/OAuth support and the client's reload/reconnect instructions.
• Unauthorized: reconnect if authorization expired or was revoked; finish browser sign-in with the intended account.
• Permission denied: inspect both the connection scopes and current project permissions.
• Revision conflict: reread and reconcile the resource before another change. Do not overwrite someone else's update.
• Rate or publishing limit: respect the returned retry instructions and account allowances; MCP does not bypass them.

MCP setup questions

Next, explore the Roblox key-system workflow, current allowances and v2 changes. Client configuration examples come from the same recipe source as the private setup screen; completing account setup still requires your browser consent.