
Ces 5 outils + Claude Code = 9.800€/mois en 2026 (spécial débutant)
Audio Summary
AI Summary
This video demonstrates the exact tools and methods used to create custom AI systems for small and medium-sized French businesses, which are then sold for €3,000 to €10,000. The approach doesn't require development skills, a common misconception. The techniques are based on experience from Wolf, a company that has helped over 300 individuals implement these tools and methods, leading to sales in various sectors like dental practices, construction, plumbing, and driving schools.
A concrete example is Bastien, a former employee who transitioned into creating AI business applications for the construction and energy sectors. Using five specific tools and Claude Code, Bastien secured a €9,800 sale, demonstrating the potential for significant monthly income.
The current landscape, in 2026, shows AI's omnipresence. While many are focused on general AI tools like ChatGPT, a largely overlooked market exists: French SMEs that need AI for survival, not just because it's "cool." These businesses are often unaware of where to start. For instance, a baker isn't interested in ChatGPT but would readily invest in an AI tool that automates catering order qualification and quote generation, as it addresses repetitive, time-consuming tasks that reduce profit margins.
To build such a tool, six tools are needed, though one, Claude Code, is likely already familiar. Most of these tools are free, requiring only a minimal subscription of around €20 to get started. Bastien, a Wolf member, was a salaried employee in construction less than a year ago. Frustrated with his financial situation, he decided to act instead of just consuming content. Today, he sells custom AI tools for €9,800 upfront, plus €300 monthly, leveraging his industry knowledge to identify and solve specific business problems. Despite his limited programming experience, beyond basic Claude Code, he learned these tools and methods like anyone else. This is a real-world case, not a theoretical exercise, with other sales ranging from €2,000 to €6,000.
The six tools are:
1. **Claude:** This acts as the lead developer. Before writing any code, Claude is used to explain the client's problem and generate a Product Requirement Document (PRD). A PRD is a detailed, literary-style specification document, akin to a project brief, used in tech to outline the tool to be developed. Historically, creating a PRD was time-consuming and required project management expertise. Now, Claude generates it, forming the project's root document.
After the PRD is validated, Claude generates the `cloud.md` file. This is a permanent context file that Claude Code reads for every request, containing necessary tools, conventions, commands, business rules, and a summarized version of the PRD as a prompt. Claude also challenges the idea, identifies potential issues, suggests security measures, outlines the technical architecture, and generates prompts for Claude Code, including system prompts for the tool itself. In essence, Claude is the right-hand man, creating while the user acts as the project manager, providing direction.
For example, to create an SEO analysis software for a WordPress CMS, the user would ask Claude to generate a PRD detailing desired functionalities. Claude then asks clarifying questions, and based on the answers, produces a comprehensive PRD outlining the problem, solution, product positioning, objectives, target users, usage scenarios, features, analysis requirements, and expected outcomes.
Once the PRD is approved, Claude is instructed to generate the `cloud.md` file and all necessary prompts for Claude Code. This `cloud.md` is placed within the Claude Code project. Claude also provides step-by-step prompts for Claude Code, guiding the development process. Claude acts as a continuous assistant, facilitating iterative development.
2. **Claude Code:** This is the developer. Once the PRD and `cloud.md` are ready, Claude Code is used to build the tool. The user provides the PRD and `cloud.md` in the terminal and instructs Claude Code to build based on the generated prompts. Claude Code then writes code, creates files, performs tests, and corrects errors. The user merely validates and observes the output.
Contrary to popular belief, development skills are not required. The user needs to understand file structures, interpret error messages, and communicate effectively with Claude Code. If unsure, Claude can explain what's happening within Claude Code. This approach allows for delivering in three days what a development agency might take two years and €30,000 to produce. This paradigm shift, similar to the early days of dropshipping, is not yet widely understood.
3. **GitHub:** This serves as the code's vault. Every action within Claude Code generates code and files stored locally on the user's computer. GitHub provides a centralized, cloud-based storage for this code, preventing data loss if the local machine fails. It also allows reverting to previous versions if Claude Code introduces bugs. Furthermore, GitHub facilitates collaborative development, acting as a central hub for teams. It's the gateway to other tools, especially Vercel. GitHub is crucial because it acts as the nerve center, enabling Claude Code to interact with and update the project, and allowing automatic communication with connected tools for updates, security, and more.
4. **Vercel:** This is the storefront, making the tool accessible to the client. While code on GitHub or a local machine is useful, it's not a usable product for a client who paid €9,800. Vercel brings the tool online via a URL. Four key aspects of Vercel are:
* **Custom Domain:** Initially, the tool might have a generic URL (e.g., `toolbastien.vercel.app`). Vercel allows for a professional custom domain (e.g., `app.bastien-plomberie.fr`), managing the domain and generating SSL certificates for security.
* **Continuous Deployment:** If a client requests a modification or new feature, the user makes changes in Claude Code, which updates GitHub. GitHub then signals Vercel to update the live tool, all within seconds, enabling rapid client response.
* **Preview Deployments:** Vercel generates temporary, separate URLs for every change made by Claude Code. This allows testing new features or modifications without affecting the live version used by the client, preventing disruptions.
* **Encrypted Environment Variables:** AI tools require API keys (e.g., for Claude, Anthropic, ChatGPT) and other credentials (e.g., Superbase keys) to function and connect to client systems. These keys are sensitive, like credit card codes. Vercel securely stores and injects these secret keys into the code in an encrypted manner, without them appearing directly in the code, providing a professional, free, and easy-to-implement layer of security.
5. **Superbase:** This is the tool's brain and memory, also offering a free tier for up to three projects. Superbase acts as the database, storing critical data like leads, quotes, statuses, and historical information. In five clicks, guided by Claude, users can create storage tables and define data to be stored. Claude handles the connection to Superbase automatically.
A significant bonus is Superbase's authentication management. For applications with multiple users (e.g., a manager and sales team), Superbase can create separate authentication pages, ensuring each user has appropriate access.
The true game-changer with Superbase, explaining why Bastien earns €9,800, is its ability to provide AI with context. Unlike ChatGPT, which relies on general knowledge, Superbase gives the AI a memory of the specific business. When a user interacts with the application, the AI accesses the company's entire database: history, margins, past projects, pricing, etc. For Bastien's construction project, this meant the AI could access historical data on bids, margins, and client interactions. Instead of a generic quote, the AI could generate a highly contextualized quote, stating, "This lead is from Chalon, we've done 12 projects for them, with a 42% margin. Here's a quote aligned with the last three deals." This significantly increases efficiency, precision, and contextual relevance, reducing errors and saving time.
6. **Stripe:** This is the payment processor. Once a client agrees to the service, Stripe facilitates online invoicing. It allows for setup fees (like Bastien's €9,800 delivery payment) and recurring monthly subscriptions (like his €300/month). The monthly payments cover updates, ongoing service maintenance, modifications, and new feature additions.
In summary, the methodology is a precise pipeline:
* **Claude:** Generates the PRD, `cloud.md`, and development strategy.
* **Claude Code:** Codes the tool.
* **GitHub:** Stores the code.
* **Vercel:** Deploys the tool online.
* **Superbase:** Stores data and provides AI memory.
* **Stripe:** Processes payments.
By mastering these six tools in this order, one can deliver custom AI tools in 3 to 7 days, depending on complexity. This creates a significant opportunity: charging €5,000, €10,000, or even €15,000 for tools that cost only €20/month in subscriptions, plus personal time. French SMEs typically budget €3,000 to €15,000 for custom tools. One client per month at €5,000 generates €60,000 annually, without employees or offices. Two clients per month result in €10,000 monthly, a life-changing income. This opportunity is time-sensitive; prices and competition will increase over the next 12-24 months.
The main challenge is sales: finding businesses, identifying their problems, and explaining the need for these tools. Businesses quickly grasp the value because they are already paying someone to perform these tasks (e.g., secretaries for quotes, sales reps for invoicing). The AI simply automates existing, repetitive tasks, freeing up staff for more impactful work like prospecting.
A common pitfall to avoid is spending months learning the tools and building side projects without ever engaging with potential clients. Someone who sells a flawed €3,000 tool learns far more than someone building a perfect tool with no clients. The recommended protocol is:
1. Find a business.
2. Sell the tool before it exists.
3. Collect a 50% deposit via Stripe upon signing.
4. Begin construction with Claude Code.
The information and tools are now laid out. The choice is to either dismiss it or start by using Claude to create a PRD for a known business case. If no case is known, the first step is to identify problems in businesses. For specific questions on tools, client acquisition, or problem identification, direct messages on Instagram are welcome.