AI Agent (MCP)
Human & AI action gateway — govern agent tool calls with the same engine as your admin back-office
Three steps — grab a token, add the MCP config for your host, then tell your AI what to build. Transcodes governs risky tool calls with step-up MFA, RBAC, and audit logs.
Package: @bigstrider/transcodes-mcp-server · Token: TRANSCODES_TOKEN from Console · Node.js required for npx
Integration guides
Governance (agent perspective)
| Topic | What it means for agents |
|---|---|
| Step-up Auth | Biometric gate before verified MCP tools (e.g. retire member) |
| RBAC | Agent acts as the token’s member; role limits which tools run |
| Audit Logs | Every agent action is logged with tag, severity, and metadata |
Console setup and end-user flows are documented under Admin Back-office. The primitives are the same — only the actor changes (human vs AI agent).
Step 1 — Prepare your token
Option A — Setup Wizard (new projects)
When you create an Authentication cluster, the Setup Wizard finishes on Register Members and shows access tokens — click Download as CSV.
Tokens are shown only once. Save them before closing the dialog.
Option B — RBAC panel (existing projects)
Open Role-Based Access Control → Members and click Get API Token for the member the MCP server should act as.
Step 2 — Add MCP config
Pick your host: Cursor · Claude · Codex · Antigravity
Step 3 — Prompt your agent
What roles exist in my project? Show recent audit logs.
Before retiring a member, walk me through step-up verification.Your agent reads project roles, resources, and settings through MCP — with step-up MFA on verified actions.
Environment variables
| Variable | Required | Description |
|---|---|---|
TRANSCODES_TOKEN | Yes | JWT from Console; encodes organization, project, and member context |
Next: Admin Back-office · API Reference