We bring you the latest updates for Amethyst.

Connect Amethyst from the AI tools you already use

Amethyst MCP can now be used as a remote MCP server. In a supported AI client, you can configure https://mcp.amethy.st/mcp and complete Amethyst authorization in the browser.

Previously, using MCP meant downloading the package, running a local server, and matching configuration details for each client. Remote MCP removes much of that setup, making it easier to call Amethyst data and analytics capabilities from the AI tools you already use.

Start with a URL and browser authorization

The Remote MCP endpoint is https://mcp.amethy.st/mcp. After you register this URL in your AI client, the client opens the Amethyst authorization flow. Once authorization is complete, the client can work with your workspace through Amethyst MCP.

The Remote MCP setup guide includes configuration examples for Claude custom connectors. If your team already uses a specific AI client, start there to confirm the right connection path.

Local MCP is still available

Remote MCP is now the recommended setup for clients that support remote servers, while Local MCP remains available for environments that need a local stdio server. We reorganized the MCP documentation into an overview page, Remote MCP setup, Local MCP setup, and a supported tools list.

That gives you a clearer choice: use Remote MCP for new setups, or keep Local MCP when you need a local server or have client-side constraints.

The foundation for more client support is in place

We added support for the scope and resource parameters used by Claude.ai Remote MCP authorization. We are also preparing support for the OAuth callback shape used by Codex. We also automated Amethyst MCP Cloud Run deployment so the Remote MCP server can be updated more reliably. A transport close issue during Remote MCP session termination has also been fixed to improve connection stability.