Pascal EK Loui · 16 Claude agents in production · 100+ skills · MCP · Agent SDK · 2 SaaS platforms built solo · Fixed price from €3,500 excl. VAT · Paris & fully remote.
A freelance Claude Code expert designs agentic systems and puts them into production, rather than just generating code faster: autonomous agents, MCP connectors to your tools, persistent skills, integration into your existing pipelines, and a handover so your teams can run the system without them.
Claude Code is marketed as a coding assistant. That is the visible part, and the least interesting one. Technically, it is an agentic environment: it reads an entire repository, runs commands, calls external tools via the Model Context Protocol and chains steps together autonomously until it reaches a result.
As a result, the most profitable uses are not the ones you would expect. Writing a function faster saves minutes. Automating a complete workflow (migrating a repository, systematic review, generating and maintaining documentation, research synthesis, an end-to-end business pipeline) saves weeks. What separates the two is not mastery of the tool. It is knowing what to hand over to an agent.
Mixing the two up gets expensive when you hire. Here is where the line falls.
| Criterion | Developer who uses AI | Claude Code expert |
|---|---|---|
| What gets faster | Their own work | Workflows that run without them |
| Unit of delivery | Code | A system with its guardrails |
| Error handling | They reread and fix | Checkpoints, error recovery and fallbacks designed upfront |
| Costs & latency | Rarely measured | Instrumented: model choice, task breakdown, budget per task |
| Connection to systems | Copy and paste | MCP connectors to databases, APIs and business applications |
| After the engagement | The knowledge leaves with them | Skills and documentation, the team takes over |
Proof, not promises: 16 Claude agents in production, 100+ reusable skills, and 2 SaaS platforms built solo on this stack: sceniq.studio (5 creative video agents, brief → reels in 48h) and tondroit.fr (legal RAG across 11 domains, sourced answers). Not a demo POC: live products with real users.
72 hours to find out whether a use case holds up: scoping, an agent in production on your data, real-world testing, costed recommendations. You decide whether to commit the scale-up budget after seeing the result, not before.
Embedded in your team to take agents from demo to production: architecture, MCP connectors, skills, monitoring, then handover. Scoped quote.
A short 21-hour programme, Qualiopi-certified and eligible for OPCO funding. Every participant leaves with a working agent, not a slide deck.
Always fixed price, never a day rate. A day rate puts the risk on the client: if the engagement drifts, they pay more. A fixed price puts it on me. It takes more work to scope, and that is exactly why the scoping is done seriously. Full details are on my 2026 rates and day rate page.
A provider who accepts everything has no method. I don't take engagements where the goal is to show off AI without a real use case, where the agent has to decide alone, without oversight, on matters with human or regulatory consequences, or where the real request is to cut jobs and have a consultant take the blame. I say so at the scoping stage, not on the invoice.
A specialist who designs, deploys and scales systems built with Claude Code and the Anthropic ecosystem: autonomous agents, MCP connectors, persistent skills, the Agent SDK, integration into existing pipelines. The job goes beyond generating code faster: they design the agents' architecture, their checkpoints and their handover.
In France, this profile is mostly billed by day rate. I work exclusively on a fixed-price basis: a 3-day Validation Sprint from €3,500 excl. VAT, and a 3- to 6-month scale-up engagement on a scoped quote. You pay for a delivered result, not for days spent.
No. It is an agentic environment: it reads a repository, runs commands, calls tools via MCP and chains autonomous steps together. The most profitable uses are not code generation but automating complete workflows: migration, review, testing, documentation, business pipelines.
A map of the tasks that can genuinely be automated, agents with their guardrails, MCP connectors to your systems, persistent skills, cost and error monitoring, and handover documentation so your teams can run the system without me.
A developer who uses AI speeds up their own work. A Claude Code expert designs systems that work without them: multi-agent orchestration, failure handling, cost and latency control, and above all deciding what should not be automated. The rare skill is architectural.
Based in Paris, Île-de-France. On site for scoping and critical workshops, fully remote for delivery. Engagements in France, Europe, the UK, the US and Asia. Reply within 48 hours.
A free 45-minute strategy call to start. We look at whether your use case holds up, and I will tell you straight if it doesn't. Reply within 48 hours.
Or by email: e.loui@uxdesignparis.fr · Phone: +33 6 35 31 50 35