Complete Guide
Your AI secretary for admin authentication, operations & management
Three steps is all you need — grab a token, add the MCP config, then tell your AI what to build — Transcodes becomes your secretary — managing roles, members, audit logs, and settings through AI Agents (Cursor, Claude Desktop, OpenAI Codex)
Package: @bigstrider/transcodes-mcp-server · You need: a Transcodes token (asdfasdfa...) from Console — set it as TRANSCODES_TOKEN (carries project and member context) · Node.js required for npx
Step 1 — Prepare Your Token
Your Transcodes token is what lets the MCP server read and modify your project (members, roles, audit logs, settings)
There are two ways to get one — pick whichever applies to you
Option A — Grab it from the Setup Wizard (recommended for new clusters)
When you create an Authentication or PWA cluster, the Setup Wizard finishes on the Register Members step and shows the access tokens for the members you just registered — click Download as CSV to save them

These tokens are shown only once — if you close this dialog without downloading or copying them, you’ll need to issue a new one from the RBAC panel (see Option B below)
Option B — Issue a new one from the RBAC panel
Already have a project but no token (lost it, rotated it, or never downloaded the CSV)? Open the project, go to Role-Based Access Control → Members, and re-issue an API token for the member you want the MCP server to act as

In the members table, click Get API Token on the row of the member you’ll use — a fresh token is generated and shown; copy it right away

Treat the token like a password — it encodes your organization, project, and member context; anyone holding it can act as that member through MCP Store it in a secret manager and never commit it to git
Keep the token handy — you’ll paste it as TRANSCODES_TOKEN in the next step
Step 2 — Add MCP Server Config
Cursor
Open Cursor Settings
- macOS: ⇧⌘J or menu bar → Cursor → Cursor Settings
- Windows/Linux: Ctrl+Shift+J or File → Preferences → Cursor Settings

Go to Tools & MCP → New MCP Server
In the left sidebar, click Tools & MCP — then click New MCP Server to add a new server

Configure mcp.json
This opens the mcp.json file:
- macOS/Linux:
~/.cursor/mcp.json - Windows:
%USERPROFILE%\.cursor\mcp.json
Add the transcodes block as shown below — if the file already has other servers, merge — don’t overwrite
{
"mcpServers": {
"transcodes": {
"command": "npx",
"args": ["-y", "@bigstrider/transcodes-mcp-server"],
"env": {
"TRANSCODES_TOKEN": "JWT_TOKEN..."
}
}
}
}
Save the file, then completely quit Cursor and relaunch it — a simple reload may not pick up MCP changes. After restarting, go back to Tools & MCP and confirm transcodes-mcp appears with tools enabled
Step 3 — Type Your Prompt
Once the MCP server is connected, just tell your AI what you need
Implement login with magic link, passkey, and TOTP,
add step-up authentication for sensitive actions,
track and store user activity logs with details like
timestamp, IP, and user agent, and include a personal
console where users can manage their authentication
methods and review their activity historyYour AI secretary reads your project’s roles, resources, and settings through MCP — then generates the right code automatically — no Console visits — no manual config
You can also delegate management tasks: “What roles exist in my project?”, “Show recent audit logs”, “Create a new member” — your secretary handles it through MCP so you never have to open the Console
Environment Variables
| Variable | Required | Description |
|---|---|---|
TRANSCODES_TOKEN | Yes | Transcodes token JWT format asdfasdf... — issued from the Console ; encodes organization, project, and member context for MCP tools |
Create or copy the token in the Console, paste it as TRANSCODES_TOKEN, save the config, then fully restart the host app so the MCP process reloads the environment
Next: LLM Context · Quick Integration · API Reference




