Model provider · in regular use
Claude agency in Nice
Anthropic's models, known for long documents and instruction following.
What we do with it
Claude at our clients, concretely
On LLM Monitor, Claude is one of the models queried to measure how visible a brand is in AI answers: chat models, Claude's web search and batch processing. We also use it in our own development work, with systematic human review before delivery.
The limits
When we recommend something else
The provider is chosen during scoping, on data sensitivity and where the model runs — never by preference.
Our work
Projects we delivered with Claude
LLM Monitor
SaaS that measures how visible brands are in AI answers.
Claude: Models queried, Claude web search and batch processing.
Read the LLM Monitor case study
Method
How we bring Claude into a project
We use Claude in two ways. In LLM Monitor, the service that measures how brands appear in AI answers, Anthropic's models are queried alongside those from OpenAI, Mistral, Gemini and Perplexity. Calls go through LangChain, which puts every provider behind the same interface: the question asked, the expected format and the way answers are stored stay identical, and only the model changes. There we use the chat models, Claude's web search and Anthropic's batch processing, which groups requests that do not need an immediate answer.
For a client, integration starts with a precise use case: which documents, what output, who reviews it. We compare Claude with other models on a sample of your own content, then pin the chosen version in the configuration so its behaviour does not change without you knowing.
We also use Claude in our own development work, to review code or draft tests. Nothing it produces is delivered without a developer reviewing it. The Anthropic account for a client project is opened in your name, with separate keys for test and production, stored outside the code.
For your business
What it changes in practice
For a small business, Claude's value shows on tasks where writing quality and following instructions matter: summarising a long document, preparing a reply that someone will review, rewording content to precise rules. The time saved is on the first draft, not on approval, which stays with a person. That is why we start no AI project without a use case costed in hours saved per week: if the gain does not hold up once review is counted, the project is not worth doing.
The trade-off is the same as with any American provider: your data passes through a third party under a different legal framework, and you depend on its commercial decisions. So the decision is made process by process, according to what the documents contain. In LLM Monitor, having several providers behind one interface is a requirement of the product; for you, it is what lets Claude be replaced by another model without rewriting the application.
Decision guide
Should you choose Claude?
Claude or another model for our use case?
We do not decide on a published leaderboard or on preference. We prepare a sample of your real documents, with the expected results, and put it to Claude and other models under the same conditions. The choice then rests on the accuracy of the answers, the cost per task and where the data is processed. Claude sometimes comes out ahead on writing while a cheaper model is perfectly adequate for simple extraction.
Claude through the API or subscriptions for our staff?
A subscription lets each person write or analyse by hand in a chat window. The API lets your software call the model automatically, on every document received and always with the same instructions. The two are not mutually exclusive. If the need is occasional and varied, a subscription is often enough; if the same task repeats every day, an API integration becomes worthwhile, provided the gain has been costed.
Can Claude handle our sensitive data?
That depends on what the data is. Anthropic is an American provider: for health data, files under professional secrecy or strategic information, we favour a model run within the European Union or on your own infrastructure. For less sensitive content, Claude remains an option after minimisation: anything the processing does not need is removed or pseudonymised before sending. We review the contractual terms in force with you during the project.
Client questions
What clients ask us about Claude
Do you actually use Claude in your projects?
Yes. LLM Monitor, our service measuring how brands appear in AI answers, queries Anthropic's models alongside those from OpenAI, Mistral, Gemini and Perplexity, including Claude's web search and batch processing. We also use Claude in our own development work, with a person systematically reviewing the output before anything is delivered.
Is code written with Claude's help reliable?
It is as reliable as the review it goes through. Claude helps us review code or draft tests, but a developer checks everything produced before it is merged, just as with hand-written code. Automated tests and code review stay exactly the same. And whatever tool is used, the code delivered belongs to you.
How much does using Claude cost?
The bill depends on the model chosen and the volume of text sent and returned. Prices change, so we do not quote them here. We estimate the cost from your actual volumes before development starts, and a spending cap is configured. For tasks that can wait, Anthropic's batch processing lets requests be grouped together.
What if we want to switch model later?
We isolate model calls in a dedicated part of the code, keeping prompts and output formats separate from everything else. That is how LLM Monitor works, with several providers side by side behind one interface. Replacing Claude with another model then means adapting that part and rerunning your reference documents, not rewriting the application.
Where do we start?
With a free first conversation, in which you describe the task you are thinking of handing to a model. Together we check whether the gain can be costed in hours per week and whether your data can be sent to an American provider. You then receive a written, costed action plan setting out the model we propose and the reasons for that choice.
Going further
Claude sits within our Artificial intelligence
A technology does not make a project. The method around it decides the outcome: scoping, prototypes, short cycles, and a clean way out.
Neighbouring technologies
A Claude project in mind?
The first conversation is free. We will tell you plainly whether this technology is the right one, and if not, which is.