renza

MCP & BYO-AI

Renza hosts no models and sells no inference. AI editing is bring-your-own: you connect your own Claude, GPT, or other assistant over the Model Context Protocol (MCP), and it acts on your documents using your own subscription. There are no AI credits and no token metering — on any plan, including Free.

This isn't a missing feature. It's the foundation of staying neutral: a generator can never credibly be a neutral home, and an AI vendor can never have zero inference cost. By never touching inference, Renza stays the asset and the contract that every model acts on.

The model

Your assistant  ──MCP──▶  Renza MCP server  ──▶  Renza API
(your subscription)        (intent-level tools)     (your org's data)

Your assistant connects to the Renza MCP server. The server exposes a small set of intent-level tools — not a one-to-one mirror of every API endpoint, but the jobs an agent actually needs: find or create a deck, replace a deck's content, publish, get a share link. Keeping the tool count small keeps it easy for an agent to choose the right action.

Under the hood those tools compose the same SDK and API you'd call yourself — so the MCP surface stays thin without losing power, and everything an agent does is subject to the same permissions and audit trail as a human.

A typical agent loop

  1. Your assistant generates or revises an HTML deck.
  2. It calls the Renza tool to import or upsert the deck (keyed by an external_id, so the same deck is updated rather than duplicated).
  3. Renza hosts the new version, re-anchoring any open comments.
  4. The assistant asks Renza for a share link and hands it back to you.

The artifact never round-trips through a Renza-hosted model. Generation happens in your assistant; Renza only ever hosts, versions, and permissions the result.

Authorization

The MCP server acts with the credentials you connect it with, scoped to your org. It can only do what your role allows, every action is attributable, and you can revoke access at any time — the same governance that applies to the API and CLI.

Availability

BYO-AI is the canonical model across every plan. Today the API, SDK, and CLI are the surfaces for wiring an assistant to Renza — anything in the app is a command or an endpoint. The MCP server ships next; this page gets the connection details the day it does.