Debate it before you build it.

A second opinion on a software or technology decision, given by several AI models instead of one. You describe your project or your question, models from different companies (Claude, OpenAI, Gemini, Mistral and others) each propose a solution and challenge the others, and an independent arbiter writes a decision report. It is used to compare solutions, test an idea before investing in it or check a quote from a provider, in ten languages.

An AI interviewer first asks about your goals, budget, team and constraints, and you validate the requirements it writes. The debaters then design their options without seeing each other, criticise each other's plans and revise them; when your answer would change the outcome, the debate pauses to ask you. The arbiter rules last. An Express debate gives a clear answer in a few minutes; a full comparison takes longer, and you get an e-mail when the debate needs your answer or the report is ready.

A single AI gives a plausible answer that nobody challenges. In Quorumind, models trained by different companies, with different blind spots, design their options separately, then look for each other's flaws, and a devil's advocate attacks every option, the favourite first. Disagreements stay visible in the report, and an option that breaks your budget, your deadline or a rule you set is marked not feasible.

For anyone who has to make or defend a technical decision, developer or not: business owners, freelancers and consultants, web agencies preparing a proposal, and teams comparing software or planning a migration. The report can be rewritten for a technical lead, an executive or a non-technical client.

In the cloud, each debate has a fixed price, AI models included, shown before it starts: €3 for Express, €7 for Standard (compare solutions, challenge an idea or audit a document) and €15 for Deep. Credits start at €10. Monthly plans include debates of any tier (a Deep debate counts as two): Solo at €49 a month for 6 debates and one person, Team at €199 for 25 debates and up to 5 people, Agency at €499 for 60 debates and any number of people. Self-hosted, you pay your AI provider directly for what your debates use, under a cap you set, or nothing with models running on your own machine.

A recommendation, the feasibility of each option, an interactive weighted matrix, a debate map, risks with mitigations, estimated build and monthly costs, an architecture diagram, a roadmap and claims marked verified, estimated or assumed. It can be rewritten for a CTO, an executive or a non-technical client, and exported as Markdown, PDF, Word or ADR.

In the cloud, Claude, OpenAI, Gemini and Mistral models, among others, through OpenRouter, with no account of your own at these providers. Self-hosted, the providers you connect with your own accounts, and local models through Ollama. You choose the model for each role: interviewer, debaters, devil's advocate and arbiter, and you can mix vendors.

Yes, to the providers of the models that play the roles: in the cloud, through OpenRouter; self-hosted, only to the providers you configure. An option masks e-mails, phone and card numbers, IBANs, IP addresses and keys before sending. Self-hosted in fully local mode with Ollama, nothing leaves your machine. You can export or delete your data and set how long sessions are kept. Quorumind does not use your debates to train models, and nobody at Quorumind reads them unless you share them, ask us to, or the law requires it.

Yes, with conditions. Self-hosted with local models, nothing leaves your machine. Otherwise, the European models only option has every role played by a model from Mistral AI, a French company, on its servers in the EU, but the requests reach it through OpenRouter, a US company: debate content stays in Europe only when OpenRouter's EU region is used, and even then OpenRouter may store account and billing metadata in the United States. In the cloud, the service and its data are hosted in France, and payments go through Stripe, which may transfer data to the United States. Each report says which route was taken. Owners and admins can turn the option on by default, or make it required, for their whole workspace.

The self-hosted version is free and open source, under the AGPL v3: you install it on your own server with Docker and pay your AI provider directly for what your debates use, or nothing with models running on your own machine. The cloud version is paid: a fixed price per debate from prepaid credits, for one person, or a monthly plan with debates included, for one person or a whole team.

A plan renews every month until it is cancelled. An owner or an admin of the workspace cancels it at any time from the Billing page, through the "Invoices and payment methods" link, which opens Stripe's customer portal. The cancellation takes effect at the end of the current month, and the plan stays usable until then; the month already started is not refunded. Credits, for their part, never expire.

Yes. Quorumind runs as an MCP server, so Claude Code, Codex or Cursor can start a debate before coding. Every report also exports as a hand-off pack with CLAUDE.md, AGENTS.md and PLAN.md.