---
title: "AI agent setup: Claude Code | MS WebX"
description: "We set up Claude Code for development teams: rules, guardrails and quality gates, tested on your own repository."
canonical: https://mswebx.com/ai-agent-setup
language: en
robots: index, follow
og_type: website
type: ai-agent-setup
language_alternates:
  en: https://mswebx.com/ai-agent-setup.md
  cs: https://mswebx.com/cs/nastaveni-ai-agentu.md
  de: https://mswebx.com/de/ki-agenten-setup.md
  es: https://mswebx.com/es/configuracion-agentes-ia.md
  fr: https://mswebx.com/fr/mise-en-place-agents-ia.md
  hu: https://mswebx.com/hu/mi-ugynokok-beallitasa.md
  it: https://mswebx.com/it/configurazione-agenti-ia.md
  nl: https://mswebx.com/nl/ai-agents-inrichten.md
  pl: https://mswebx.com/pl/konfiguracja-agentow-ai.md
  pt: https://mswebx.com/pt/configuracao-agentes-ia.md
  ru: https://mswebx.com/ru/настройка-ии-агентов.md
  tr: https://mswebx.com/tr/yapay-zeka-ajani-kurulumu.md
  zh: https://mswebx.com/zh/ai智能体搭建.md
  ja: https://mswebx.com/ja/aiエージェント導入.md
  ar: https://mswebx.com/ar/إعداد-وكلاء-الذكاء-الاصطناعي.md
  ko: https://mswebx.com/ko/ai-에이전트-구축.md
  pt-br: https://mswebx.com/pt-br/configuracao-agentes-ia.md
  hi: https://mswebx.com/hi/एआई-एजेंट-सेटअप.md
content_hash: sha256:dfcf1f8a8511acf5cc3b87188829094eb621698c1c3c832be730e6ac4ce8bddd
generated: 2026-11-17
---

> **Machine-readable Markdown rendering.** This is a Markdown representation of a web page,
> served for AI agents and crawlers as a token-efficient alternative to the full HTML.
> Full interactive HTML version: https://mswebx.com/ai-agent-setup
> Canonical URL: https://mswebx.com/ai-agent-setup
> Available languages: en, cs, de, es, fr, hu, it, nl, pl, pt, ru, tr, zh, ja, ar, ko, pt-br, hi
> Change detection: `content_hash` is the SHA-256 of this document's Markdown body — everything
> below this block, whitespace-trimmed, excluding the front-matter. Compare it to detect a real
> content change; `generated` moves on every render and is not a change signal.

AI agent setup

#  Claude Code that works *on your real codebase*

We set up AI coding agents for development teams: a rules file the agent follows, guardrails against dangerous actions, quality gates before anything merges, memory and tool connections. All of it is tested on your repository, written down and handed over to your team. Fixed scope, billed per day.

 [Book a free call](#book) [See what you get](#deliverables)

No promised productivity numbers. A fixed scope and written deliverables.

     your-repo — AI agent

1. Task: add invoice export to the billing module
2. Rules loaded *CLAUDE.md · project rules · skills*
3. Plan written, decision recorded
4. Writing tests first, then code
5. Guardrail blocked `git reset --hard` *hook*
6. Running quality gates
7. **tests** **types** **lint** **security**
8. Ready for review: one branch, one change

  Rules Guardrails Quality gates Tool connections

  years of software engineering

 20+

 specialised AI agents

 27

 knowledge modules

 73

 own MCP servers

 5

 custom code rules

 215

 Cloudflare agent readiness

 Level 5

 Our own setup, built on Claude Code. Agents access production data read-only; guardrails prevent data loss.

 What you get

## Six things we put in place

Everything is set up on your own repository, tested on real tasks and written down.

-

### Rules the agent follows

    A project rules file that tells the agent how your codebase works: conventions, architecture, commands and the things it must never do.
-

### Guardrails against dangerous actions

    Hooks that stop destructive commands before they run, such as a forced reset or a delete on production data.
-

### Quality gates before anything merges

    Tests, type checks, linters and security checks run automatically, so the agent’s work arrives in review already checked.
-

### Memory and reusable skills

    Project knowledge and recurring tasks are written down once, so neither the agent nor your team starts from zero in every session.
-

### Tool connections over MCP

    The agent gets access to the tools your team already uses, through MCP servers with exactly the permissions you choose.
-

### Playbook and handover

    A written playbook for your team and a handover session. The configuration lives in your repository, so it stays with you.

 Packages

## Pick the scope that fits your team

Every package has a fixed scope and is billed per day. Start small and extend when you are ready.

-

### Agent readiness review

    We read your repository and your workflow and tell you in writing what stands between your team and a reliable agent, with prioritised next steps.
-

### Agent setup day

    One working day on your repository: rules file, guardrails and the first quality gates, tested on a real task and written down.
-

### Team sprint

    Several days for a whole team: rules, guardrails, quality gates, memory, tool connections, a playbook and a handover session. Includes 30 days of questions afterwards.

    Includes a written training record for EU AI Act Art. 4 (AI literacy).
-

### Monthly support

    A fixed monthly retainer for questions, rule updates and reviews while your agents and your codebase keep changing.

Fixed price, quoted in the free call

 How it works

## From first call to handover in four steps

1.

### Free 30-minute call

    You tell us about your project and your goal. We tell you honestly whether and how we can help. No obligation.
2.

### Readiness review

    We read your repository and workflow and write down what to change first, so the scope is clear before work starts.
3.

### Fixed-price days on your repository

    We set up rules, guardrails and quality gates on your own repository and test them on real tasks together with your developers.
4.

### Handover

    You get the playbook, the configuration in your repository and a handover session with your team. Everything stays yours.

 How we build

## AI agents write the code. A strict contract keeps it right.

The same loop runs our game engine and goes into your repository when we set up your agents.

1.

### Plan

    The owner owns the plan. Features go PRD → tasks → process.
2.

### Decide

    Every decision is written down as an ADR before code exists.
3.

### Lane

    One change, one lane. Specialist agents, one lead.
4.

### Gates

    Tests and a lint wall that only gets stricter. A new gate must first fail once.
5.

### Merge

    Only green work lands. Agents cannot loosen their own rules.

 Questions

## Frequently asked questions

Where does our code go?

We work on your own repository, in your accounts and with the access you grant us. The configuration lives in your repository and stays with you. Which data an AI provider processes depends on the plan and settings you choose with that provider; we go through them with you before we start. We keep no copies of your code. On request we sign an NDA and a data processing agreement.

Will it work for our stack?

In the free call we look at your language, framework, build and tests and tell you honestly whether a setup is worth it for you. The readiness review puts that assessment in writing before you commit to more.

Do you guarantee results?

No. We guarantee a fixed scope and written deliverables, not business outcomes such as rankings, traffic or productivity gains. Those depend on factors we do not control.

Which AI tools do you set up?

Our focus is Claude Code. The same building blocks apply to other coding agents; in the free call we tell you where we can help and where we cannot.

Who does the work?

We do, in-house. We have more than 20 years of full-stack experience and more than 4 years of working with AI. Our own setup runs 27 specialised AI agents with 73 knowledge modules and 5 MCP servers of our own, and nothing ships without passing our automated quality gates.

How quickly do you reply?

We reply to every request by email within one business day.

Does this help with the EU AI Act?

Article 4 of the EU AI Act asks companies that use AI to ensure a sufficient level of AI literacy among their staff. The Team Sprint includes a written record of the training your team received. This is not legal advice.

 Free 30-minute call

## Tell us about your team

A few details are enough. We reply by email and suggest a time for the call.

We reply within one business day.

Prefer email? [hello@mswebx.com](mailto:hello@mswebx.com "hello@mswebx.com")

  Your name

 Work email

 Company

 Developers in your team  Please choose 1-5 6-15 16-50 51+

 Stack and repositories

 AI coding tool today  Please choose Nothing yet Claude Code GitHub Copilot Cursor Something else

 Preferred time for the call (optional)

 What hurts most? (optional)

        I agree that my details are used to answer this request. [Privacy](/privacy "Privacy Policy")  Send request

### Thank you, your request has arrived

A confirmation is on its way to your inbox.

 More services

## What else we do

### AI integration

AI features inside your product and workflows, with typed contracts and audit logging behind them.

 [Learn more](/ai-integration "AI Integration")

### Websites

Server-rendered, fast and multilingual. We run our own sites in 18 languages.

 [Learn more](/websites "Websites")

### SEO & GEO

Rankings in search engines and citations in AI assistants: structured data, hreflang, answer-first content.

 [Learn more](/seo-geo "SEO")

## Struggling with AI? Let’s talk.

Book a free 30-minute call. We reply within one business day.

 [Contact us](/contact "Contact")
