BrainSoup
BrainSoup is a native Windows application for building teams of specialised AI agents that collaborate through plain conversation. It keeps data in a local database, supports local LLMs, and starts at 5 USD per month.
What is BrainSoup?
BrainSoup is a desktop application for Windows 10 and above, published by the French software house Nurgo Software. Its premise is that a team of narrow specialists beats one generalist: rather than talk to a single assistant, you assemble a set of custom AI agents, each built for a particular job, and let them work together.
The interface deliberately looks like a messaging app. You give instructions in ordinary language, as you would to a colleague, and agents answer in a thread you can search. When a task is too big for one of them, several meet in a goal-oriented chat room, each with its own skills and its own access to data, which keeps roles tight and limits what any single agent can reach. A sandboxed file system governs how agents and people exchange files.
Underneath, BrainSoup can call ChatGPT or Mistral AI, or a model running locally through a tool such as Ollama. Semantic Kernel technology gives the agents an internal sense of time and self, so they remember you and recall what was said in earlier conversations. You widen what they know by dropping in your own documents, or simply by talking to them over time.
Agents also act rather than only answer. Built-in tools let them calculate, browse the web, send emails, call APIs, generate and edit images and video, and run scripts; ChatGPT plugins extend that further, and you can declare custom tools pointing at your own programs and commands. Extension is deliberately low-code: BrainSoup talks to your environment through plain text files and scripts in any language, with no API, no SDK and no proprietary language to learn. Agents handle images and audio as well as text, and simpler models are made multimodal by pairing them with others.
Data stays in a local database on your machine and there is no cloud storage. A built-in expert agent called Doc-E, itself built with BrainSoup, walks you through setup and script writing and doubles as a demonstration of what you could build. The product is sold as a no-commitment subscription from 5 USD per month, monthly credits included.
What it does
- Create specialised AI agents, one per job to be done
- Have several agents collaborate in goal-oriented chat rooms
- Drive automations by talking to agents in plain language
- Feed agents your own documents to shape what they know
- Run everything against a local language model instead of a remote service
- Let agents browse the web, send emails, call APIs, generate images and video, and run scripts
- Trigger agents from events you define and wire them to your own programs
When to use BrainSoup / When not to
A quick filter to help you decide if BrainSoup is the right fit.
When to use BrainSoup
- Windows users who want to automate recurring desk work without writing code
- Privacy-conscious professionals who need their working data to stay on their own machine
- Developers and scripters who would rather extend a tool with text files and scripts than learn an SDK
- Solo operators and small teams looking for a low entry price rather than per-seat licensing
- Curious learners who want hands-on practice building and orchestrating AI agents
When not to use BrainSoup
- macOS, Linux and mobile users, since the application runs only on Windows 10 and above
- Teams needing a browser-based or hosted deployment, as no web version exists
- Developers hoping to embed the product through a public API or SDK, which it deliberately does not offer
- Organisations whose procurement requires a product licence agreement, a DPA or a documented GDPR framework
- Anyone wanting to try the software at length for free, since a paid subscription is required from the start
How to use BrainSoup
A typical end-to-end flow, from setup to results.
- Check that you are on Windows 10 or later, the only platform supported today
- Download the BrainSoup.msi installer from the product page and run it
- Take out a subscription, which is required to use the software at all
- Create your first agent and describe in plain language what it is meant to do
- Talk to it in the messaging interface and refine its behaviour from its answers
- Drop your own documents into the Documents folder to widen what it knows
- Add more specialists and bring them together in a goal-oriented chat room
- Give agents tools, from the built-in set or as custom tools pointing at your own scripts
- Use the Context, Events and Logs folders to feed live data, trigger agents and trace what happened
- Ask Doc-E, the built-in expert agent, whenever setup or an extension script blocks you
Pros & Cons
Pros
- Working data stays in a local database on your own machine, with no cloud storage
- A local language model can be used through Ollama, so nothing need leave the computer at all
- The publisher states customer data is never used for training and that contracts with third-party services forbid it
- Low entry price at 5 USD per month, with no commitment and cancellation at any time
- Unusually frank about its economics: third-party AI costs are passed on without margin
- Extending it needs no API or SDK, only text files and scripts in whatever language you already use
- Backed by a knowledge base of 69 product-specific articles and a publisher trading since 2009
Cons
- Windows only: no macOS, Linux, iOS or Android build, and no web version either
- No free plan and no documented free trial; the subscription is required from the first use
- No pricing page at all — the price lives in one knowledge base article, and the product's pricing URL returns a 404
- No licence agreement is published for BrainSoup; the Licences page covers only two older products
- Not one mention of GDPR anywhere, no retention period, no DPA and no described data rights
- The language model providers that actually process your prompts are never named as subprocessors
- Running costs are hard to predict, since credits are consumed according to the model and the volume of data
Pricing & Plans
There is no free plan: a subscription is required to use BrainSoup at all. The lowest advertised entry point is 5 USD per month, billed without commitment and cancellable at any time. That subscription includes a monthly allowance of credits, consumed each time an agent is activated, with consumption varying according to the language model used and the volume of data processed. No free trial is documented, and no annual billing option is offered.
- Subscription from 5 USD per month
- no commitment
- monthly credits included
- the wording "starting at" implies higher tiers
- but no plan grid is published
Data, GDPR & hosting
A consolidated view of how BrainSoup handles your data.
GDPR overview
There is no GDPR framework published anywhere on this site. A search across every page collected — terms of use, privacy policy, licences, knowledge base — returns not a single occurrence of the regulation. The privacy policy states no legal basis for processing, no retention period, no right of access, rectification, erasure or portability, and describes no complaint procedure. No data protection officer is named and no data processing agreement is offered. The rating of N therefore records silence, not a refusal: the publisher never claims compliance and never denies it. Two things soften the practical exposure without repairing the documentation. Nurgo Software is established in France, so the regulation applies regardless, and BrainSoup keeps working data on the user's own machine. The only privacy contact is the general company address at the foot of the policy.
Who owns the data?
BrainSoup keeps everything you create in a database held on your own machine, under the %AppData% folder, and the publisher states it deliberately avoids cloud storage. Nurgo Software declares its access limited strictly to the information needed to bill your subscription. The published terms of use claim ownership only over content posted through the website's open communication tools, such as comments and forums; they say nothing at all about the material handled inside the application. The privacy policy covers the website and, by its own wording, every Nurgo product, yet it never addresses ownership of the content the agents process.
Reuse rights
Nothing in the terms restricts what you do with the output your agents produce: the licence Nurgo Software reserves for itself applies only to content posted through the website's public communication tools, not to work carried out inside the application. Since that work stays in a local database on your machine, reuse is a practical matter rather than a question of permission. Requests sent to third-party AI services are described as anonymised and de-linked from your identity, and the publisher says its contracts with those providers forbid using them to train models. Outside the product, the site declares it shares information only with Google Analytics, Facebook Custom Audience and Twitter Tailored Audiences.
Data retention & training
Hosting summary
BrainSoup does not host your working data at all: it is written to a local database on your own machine, under the %AppData% folder on Windows, and the publisher states it deliberately avoids cloud storage. There is consequently no hosting country or region to declare for the product itself, and the site names none. What does leave the machine are the requests sent to third-party AI services such as ChatGPT or Mistral AI; the infrastructure behind those services is never identified and no jurisdiction is stated for it. Choosing a local model through Ollama removes that outbound traffic entirely. On the publisher's side, only subscription billing data is said to be accessible, again with no location given. For reference, the website itself resolves to an OVH address in Roubaix, France, but that is a network observation about the site, not a declaration about where anyone's data lives. Buyers who need a documented hosting jurisdiction will not find one here.
Things to keep in mind
Risks and trade-offs to weigh before adopting BrainSoup.
- No licence agreement is published for BrainSoup, and neither the terms of use nor the privacy policy ever addresses the product: both are generic website documents, one of which still sends disputes to American courts despite a French publisher
- The absence of any GDPR wording, retention period or deletion procedure means you hold no written commitment about how long anything is kept or how to have it removed
- Agents can browse the web, send emails, call APIs and run scripts on your machine; handing that much reach to a system driven in plain language deserves care about which agent gets which tool
- Because the database is local and there is no cloud sync, losing or wiping the machine loses the agents' accumulated memory; backups are entirely your responsibility
- Costs are open-ended by design: credits are consumed per agent activation, so an over-eager multi-agent workflow can spend well beyond what the 5 USD entry price suggests
- Delegating routine thinking to a standing team of agents is comfortable, and comfort erodes the habit of checking output; a model's mistakes arrive in the same confident tone as its good work
- The testimonials on the product page are anonymous and several describe a single week of use, while the site copyright stops at 2024 — thin signals on which to judge maturity
Setup & Integrations
Technical difficulty
Low to begin with, moderate to go far. Installation is a single Windows .msi, and the messaging-style interface needs little explanation; a built-in expert agent, Doc-E, answers setup questions and helps write extension scripts. Beyond conversation, you have to grasp the Documents, Context, Events and Logs folders that carry knowledge, live data and triggers, and custom tools mean writing scripts of your own. Pointing BrainSoup at a local model requires installing Ollama separately. No API or SDK has to be learned, which keeps the ceiling lower than the feature list suggests.
Deployment
Integrations
Behind BrainSoup
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement BrainSoup.
Frequently asked questions
Which operating systems does BrainSoup run on?
Is there a web version?
How much does it cost?
Is there a free plan or a free trial?
Is my data used to train the language models?
Where is my data stored?
Can I use a local language model?
Does BrainSoup offer an API?
What does the site say about GDPR?
Can I cancel at any time?
Should you pick BrainSoup?
BrainSoup makes an unusual offer in a market that has largely moved to the browser: a native Windows application where you assemble a small team of specialised AI agents, keep their database on your own machine, and can point the whole thing at a local model so that nothing leaves the computer. For anyone who has hesitated over sending working documents to a hosted assistant, that architecture is the main argument, and concrete statements back it — no cloud storage, publisher access limited to billing data, contracts with third-party services forbidding training on customer data.
The product is real and documented: sixty-nine knowledge base articles, a downloadable installer, and a small French publisher selling Windows software since 2009. The entry price is low at 5 USD per month, and the credit model is explained with rare candour, since third-party AI costs are passed through without margin.
What has not kept pace is the paperwork. There is no licence agreement for BrainSoup, no pricing page, and not one mention of GDPR anywhere on the site — no retention period, no data processing agreement, no described route for exercising data rights. The terms of use remain a generic template pointing disputes at American courts, which sits oddly with a French company. None of that contradicts what the software does, but a buyer with a procurement checklist will have questions the site does not answer.
The scope is narrow today as well: Windows only, one machine, no web version and no API, against a broad but undated roadmap. BrainSoup suits an individual or a small team at ease on Windows who values keeping data local more than portability or a formal contractual framework.
- Choosing a selection results in a full page refresh.
- Opens in a new window.