Stytch MCP Server

The Stytch MCP Server connects an AI assistant to your Stytch workspace. Stytch is an authentication platform that teams use to build login, single sign-on, and further access features into their applications. Instead of hunting for settings in the dashboard, you state your request in chat or in the editor, for example "What are my Stytch project IDs?" or "Upload the email templates from my git repository to Stytch". The server is operated by Stytch itself; according to the documentation the company has implemented its own product as an MCP server.

What is the Stytch MCP Server?

MCP, the Model Context Protocol, is an open standard through which AI assistants can call tools of external services. The Stytch server provides such tools so that an assistant can carry out configuration tasks in Stytch on behalf of the signed-in person. According to the documentation the address is https://mcp.stytch.dev/mcp; it is a hosted remote server, so you do not have to install or run anything locally. In the Cursor example in the documentation, the client reports 21 available tools after sign-in. The documentation does not list these tools individually; you see them in your client's tool list.

What is it particularly useful for?

The value lies in recurring configuration work. The documentation gives uploading email templates from a git repository as an example. It also suggests writing Stytch setup steps into a project's README so that an agent can carry them out directly; an example repository with Vercel is linked for this. That is handy for teams that often set up new environments: the instructions live next to the code, and the assistant works through them. Compared with a generic HTTP tool, this server knows Stytch's terms and procedures and acts within your workspace permissions.

Setup

According to the documentation, mcp.stytch.dev offers a one-click install for Cursor and VS Code. Alternatively, in Cursor you add a custom MCP server under Settings, Tools and Integrations, and put the address into the mcp.json file. Stytch then shows the note "Needs authentication"; one click starts the sign-in with which you grant the client access to your workspace. From time to time you have to sign in again. For other clients the documentation names the same address, though the steps may differ.

Permissions and security

Configuration tasks change state, so Cursor by default asks for your approval before every tool call. If you let it confirm automatically, be aware that the assistant can then change settings of an authentication service without asking. The documentation contains neither a dedicated security page nor a license statement. Check changes in a test project first before you point the server at production projects.

Costs

The server requires a Stytch workspace. According to the pricing page there is a free entry tier, usage-based plans without a base fee, and an enterprise tier with individual terms. Whether a specific plan level is needed for MCP access is not stated in the documentation; details are on Stytch's official pricing page.

Who is the Stytch MCP Server for?

It is worthwhile for development teams that already use Stytch and would rather manage configuration, templates, and project data in conversation with an assistant than by hand in the dashboard. Anyone not using Stytch gains nothing from it: the server is not a standalone sign-in solution but an access path to managing the platform. Teams should also clarify who may let the assistant loose on production projects.

A Stytch account with a workspace and an MCP client that supports remote servers, such as Cursor, VS Code, or Claude Code.

With the Stytch MCP Server, an AI assistant manages your Stytch configuration. This guide leads from setup to the first request.

Step 1: Check the prerequisites

You need a Stytch account with a workspace and an MCP client that supports remote servers, such as Cursor, VS Code, or Claude Code. Nothing has to be installed locally because the server is hosted by Stytch.

Step 2: Register the server

According to the documentation the address is https://mcp.stytch.dev/mcp. On mcp.stytch.dev there is a one-click install for Cursor and VS Code. In Claude Code the command claude mcp add --transport http stytch https://mcp.stytch.dev/mcp is enough.

claude mcp add --transport http stytch https://mcp.stytch.dev/mcp

Step 3: Set up manually in Cursor

Open the settings, choose Tools and Integrations, and add a custom MCP server. In the mcp.json file, add an entry stytch under mcpServers with the address as url. Restart Cursor if the entry does not appear.

Step 4: Sign in

After registering, Stytch shows the note "Needs authentication". One click starts the browser sign-in with which you grant the client access to your workspace. Cursor then lists the tools; in the documentation there are 21. An occasional re-authentication may be needed.

Step 5: First requests

Start read-only, for example with "What are my Stytch project IDs?". You do not have to name a tool; natural language is enough. After that you can try tasks such as uploading email templates from a git repository.

Step 6: Write setup steps into the README

The documentation recommends writing Stytch steps into your project README so that an agent can carry them out. That keeps the instructions next to the code and reusable for new environments.

Permissions and caution

Configuration tasks change state. Cursor therefore asks for approval before every tool call by default. If the agent seems to hang, it is often waiting for this confirmation. Test in a test project first before you connect production projects.

Browser sign-in with your own Stytch account; the client thereby gains access to the workspace. An occasional re-authentication may be needed.

Access to configuration and projects of the Stytch workspace within the rights of the signed-in person, including changing actions such as uploading email templates.

The server runs hosted by Stytch. The setup page gives no details on data processing; retrieved configuration data enters the AI client context and therefore reaches the model provider.

Unwanted configuration changes to an authentication service, especially with automatically approved tool calls, and passing project data to the model provider.