Your developers already use it
Each with their own settings and habits. You want shared rules and controlled costs.
Claude Code · Anthropic · Software development
Claude Code: an assistant for your developers, with limited access.
Claude Code writes, changes and tests code directly in your projects. We define what it can read, run and change, how its code is reviewed, then train your developers.
01 Understand02 Decide03 Build04 Verify05 Measure
Claude CodeAnthropic01Tasks
02Access
03Rules
04Training
When is it useful?
Sensitive code
Customer data, server access, API keys: the agent must not see or run everything.
Old code to maintain
Missing tests, no documentation, overdue updates: a well-supervised agent moves this work forward.
What we deliver
Shared settings
A shared configuration file: allowed commands, forbidden folders, network access.
Connected tools
The list of approved connectors (MCP servers, which give the agent access to other systems), each with an owner.
Review and tests
Which changes need human review, and which tests must pass before merging.
Team training
A session on your real repositories, then a follow-up after a few weeks.
For an SME: where to start
This work falls under our Engineering offer. For a small development team, a support subscription may be enough afterwards.
AI Engineering
Configuration, access rules and integration with your tools, at a fixed price on proposal.
Fixed price on proposal, or subscription from CHF 1,800/month (10 h) excl. VAT
See the offer AI Engineering →AI Adoption Support
Monthly team follow-up: new uses, settings and review of results.
from CHF 1,800/month (10 h) excl. VAT, 3-month minimum
See the offer AI Adoption Support →Tool licences are paid directly to the vendor. They are not included in our prices.
Points to watch
Keys and passwords
An agent that reads the developer’s whole machine also reads their credentials. Secrets stay out of its reach.
Unvetted connectors
An MCP server found online can read or send data. Only vetted tools are allowed.
Cost
Usage varies widely. We set up cost tracking per team.
Frequently asked questions
Is Claude Code safe for our source code?
It depends on the settings, not on the product alone. Anthropic documents permission controls; we configure them for your repositories: what the agent can read, run and send.
Claude Code or OpenAI Codex?
We compare both on your repositories and the same tasks: code quality, tests, cost and how easy they are to supervise.
Can it run through Amazon Bedrock or Google Cloud?
Yes, Anthropic documents both options. They are useful if your company already has its contracts and rules with one of those providers.
Is Numezis an Anthropic partner?
No. Numezis does not claim any Anthropic or Claude partnership or certification. We advise and implement independently, and choose the tool for your work.
A first conversation to agree the next step.
Describe your need in a few lines. We reply within two working days with the offer that fits.