Quorum in Replit, Lovable, Bolt, v0 and Base44

Add Quorum to your AI app builder in one paste, or call the API from the app you build.

There are two ways to use Quorum with an AI app builder. Add it to the builder, so the agent can convene a panel while it works. Or call it from the app you ship, so your users get the panel.

Add Quorum to the builder

All five use OAuth. Paste the server URL, sign in to Quorum and approve. You approve once.

https://www.quorum.dog/mcp

Replit

1. Open Agent · MCP servers.
2. Copy URL and paste it.
3. Sign in to Quorum and approve.

Lovable

1. Open Settings → Connectors.
2. Copy URL and paste it.
3. Sign in to Quorum and approve.

Bolt

1. Open Connectors → custom MCP.
2. Copy URL and paste it.
3. Sign in to Quorum and approve.

v0

1. Open + menu → MCPs.
2. Copy URL and paste it.
3. Sign in to Quorum and approve.

Base44

1. Open Account → MCP connections.
2. Copy URL and paste it.
3. Sign in to Quorum and approve.

Once connected, ask the agent to call list_modes. The call is free and returns the Modes you can use. Quorum exposes 13 tools.

A connected app on your plan pays real cost, with no Mode fee.

Call Quorum from your app

The API is OpenAI-compatible. Call it from your server or a backend function, where the key stays private: a request sent from a browser gets a 403 with browser_origin_not_allowed. Point baseURL at Quorum, set model to a Mode ticker, and set the timeout to 300 seconds.

import OpenAI from "openai";

const question = "Should we move billing to usage-based pricing?";

const client = new OpenAI({
  baseURL: "https://www.quorumllm.com/v1",
  apiKey: process.env.QUORUM_API_KEY,
  timeout: 300 * 1000, // milliseconds
});

const r = await client.chat.completions.create({
  model: "QRUM:STAN",
  messages: [{ role: "user", content: question }],
});

console.log(r.choices[0].message.content);

model names a Mode, in the form COMPANY:MODE. A Mode is a saved setup of which models sit in which seats. Modes stay stable while the engines under them change. API keys always bill your organization's API account.

Two things to build around:

• stream: true returns a 400 with stream_not_supported. Generated chat interfaces often stream by default, so change that call.
• A deliberation takes longer than one model call. Put it behind a button, show progress while the user waits, and set your client timeout to 300 seconds.

Call POST /v1/estimate first to decide whether a question deserves a panel. It is free.

Where it fits in an app

• The one screen that has to be right. The review step, the recommendation, the thing a user acts on.
• Behind a button. A button labelled "Get a second opinion" tells the user to expect a wait.
• As a check on your own output. Generate with a fast model, then ask the panel what is wrong with it.

Related

Modes · Escalation router · Pricing · API reference