Put Claude Code to work in your developer team, with clear rules
Agentas sets up Claude Code for companies with their own developers, from 2 to 30. One shared setup in your repositories, a change flow where your CI decides what gets merged, rules for what the AI may see, and a short training for the team. We use Claude Code in our own work every day, and we show your team how we do it.
What you gain
One setup for the whole team
A CLAUDE.md in each repository tells Claude Code how your project is built and tested. Shared permission rules and hooks decide what it may run on its own. It all lives in git, so every developer starts from the same place.
Changes go through review
Claude Code works on a branch and opens a pull request. Your tests, linting and checks in CI have to pass before anything is merged, and a developer reads the change. The AI writes faster, and the gate stays yours.
Rules for what the AI may see
In regulated companies, finance for example, we write down which code and data may go to a model, how Claude Code is connected and what is logged. Your compliance people sign off on the result.
A team that knows how to use it
A short hands-on training on your own code: how to give Claude Code a task, how to check its work and when to take over. Afterwards the setup is running and the team knows why each rule is there.
Who this is for
- Companies with their own developers, 2 to 30, that already use Git and have some automated tests or CI.
- Teams where some developers already use Claude Code on their own and the company wants one shared way of working.
- Regulated companies that need a written rule for what the AI may see before they can say yes.
- Less relevant if you have no developers in house. Then we build the software for you instead.
How we build it
We start with a short call about your team, your repositories and your rules, so we both know if this fits. If it does, the first step is a one-day workshop on one of your own repositories. We set up the shared files, connect it to your CI, agree the data rules and train the developers in the same sessions. After a few weeks we look at how it is going and adjust the rules. Follow-up and advice are billed by the hour.
It is the setup we run ourselves: a CLAUDE.md in each repository, permission rules and hooks that block what must not happen, and checks that have to pass before anything is deployed.
Read how we work →What you can count on
- You keep everything we write. Instructions, settings, hooks and data rules live in your repositories and your documents.
- We do not sell Claude licences and take no commission from Anthropic or any cloud provider. You buy access directly.
- We tell you where Claude Code is weak and where a person still has to read the code.
What businesses search for
Some of what Norwegian businesses type into Google when they need this — shown here as plain content, not hidden keywords.
Questions
Can a regulated company use Claude Code?
Often yes. It depends on how Claude Code is connected and what it is allowed to see. Under Anthropic's commercial terms, Anthropic does not train its models on code or prompts sent through Claude Code, and zero data retention is available to qualified Enterprise accounts by agreement. Claude Code can also run through Amazon Bedrock, Google Cloud or Microsoft Foundry, so the model is called through a cloud account you already use. We help you choose and write the rule. The decision is your compliance function's, and we are not lawyers.
What is a CLAUDE.md file?
A text file in the repository that Claude Code reads at the start of every session. It holds what a new colleague would need: how to build and test, the code conventions and what not to touch. Short, specific files work better than long ones.
How do we stop Claude Code from running something it should not?
Permission rules say which commands it may run without asking and which are blocked. Hooks run your own script before or after an action, for example to stop a change to a protected file. Both live in a settings file in the repository, and your admins can set organisation-wide settings that individual developers cannot override.
Does pull request review still matter when AI writes the code?
Yes. Claude Code can write and change a lot of code quickly, so the checks around it matter more. We set it up so a change has to pass your tests and a developer has to read it before it is merged. Claude Code can also review pull requests in CI as one more reader. A person still approves.
How big a team is this for?
Two to thirty developers. In a larger organisation we start with one team and let the setup spread from there.
What does it cost?
We quote a fixed price for the workshop after the first call. Follow-up and advice are billed by the hour.
Do you also work with other AI coding tools?
Our focus is Claude Code, because it is what we use every day. Much of the setup carries over: project instructions, CI as the gate and the data rules apply to any tool. We do not train teams on other tools.
Related services
AI architecture and AI architect services
AI architecture from Agentas AS: an AI architect who designs your LLM, RAG and agent systems with evaluation, security and cost control, then builds them.
Read more →AI and LLM integration into your systems
Integrate LLMs into your product or workflow — agents, retrieval (RAG), and tool-use built for production with evaluation harnesses and guardrails. Delivered by Agentas AS.
Read more →Have a project in mind?
Tell us what you are trying to do — you get an honest read on feasibility, and a fast one.
Get in touch