Fixed-price project

Claude Code setup: sub-agents, skills and MCP servers for your repo

I set up Claude Code in your repo with rules, safe permissions, sub-agents, skills and MCP servers, plus a guide for your team. It suits developers and small teams, from $250 for a base setup in 3 days.

Price
From $250
Delivery
3 days
at the Base setup tier
Packages
3 tiers

Who it is for

Who this helps

  • Developers who use Claude Code but get uneven results from one long chat
  • Small teams that want every developer to start from the same agent rules
  • Tech leads who need to control what agents may do and what they cost
  • Teams new to AI coding tools who do not want to learn it all by trial and error

Signs

Signs you need this

  • The agent forgets your project rules in each new session.
  • Every developer has a different private setup, or none at all.
  • You worry about what the agent is allowed to run or change.
  • Usage limits run out faster than the work gets done.
  • You want the agent to read your docs or database safely, but you do not know how to connect them.

Scope

What I do

  • A short rules file

    I write a CLAUDE.md for your repo with your stack, commands, code style and the things the agent must never do. The Claude Code docs say these files are context, not enforced rules, and suggest keeping each one under 200 lines. So I keep it short and put the hard limits elsewhere.

  • Safe permissions and hooks

    Claude Code lets you allow, ask or deny each kind of action. I set rules so the agent can read code and run your tests freely, and has to ask or stop before risky commands. The rules live in your repo, so the whole team shares them. In the larger packages I add hooks, which run automatically at set points, for the rules that must never break.

  • Sub-agents and cost rules

    A sub-agent is a helper with its own context window, its own tools and its own instructions. I define a few for the jobs you repeat, such as searching code, reading docs, build work or review. Cost rules send simple jobs to cheaper models, so the main chat stays focused and simple work can cost less.

  • Skills and MCP servers

    Skills are short instruction files for tasks you repeat, such as a release checklist. MCP servers give the agent safe access to docs, databases or your tools. I connect only what you need. The Claude Code docs warn that servers which fetch outside content can expose you to prompt injection, so I explain what each one can reach.

  • What I need from you

    Your stack, access to the repo I should work in, and what you want agents to do first. You use your own Claude account, so usage and bills stay with you. If your team has rules for code review or releases, send them, and I will build them in.

  • What is not included

    I do not write your product features, and I do not run your agents for you after the setup. Claude plan or API costs are paid by you. Setups for other AI coding tools are a separate talk. I do not promise a set saving in time or cost.

Packages

Pick a package

Fixed price in US dollars. Each package lists what is included.

Base setup

$250

3 days · 1 revision

One repo and one developer or a small team starting with Claude Code.

  • Rules file for one repo
  • Safe permission rules
  • 2 sub-agents
  • Written guide for your team

Full agent team

$750

7 days · 2 revisions

A team that wants agents for search, docs, build and review, with limits on risk and cost.

  • Agent team of up to 5 sub-agents
  • 3 skills
  • 2 MCP servers
  • Hooks and cost rules

Team rollout

$1,800

14 days · 3 revisions

A team that wants the setup working for every developer, with training.

  • Up to 8 agents
  • 6 skills
  • A custom MCP server
  • A CI job
  • A team training call

Process

How it works

  1. Step 1

    You tell me your stack and first jobs

    You describe the repo, the stack and the first things you want your agents to do.

  2. Step 2

    I set it up in your repo

    I add the rules, permissions, agents, skills and MCP servers, in your style and for your stack.

  3. Step 3

    I check it on a real task

    I try the setup on a small task from your code, fix what does not work, and write the guide.

  4. Step 4

    Walkthrough and handover

    I walk you through the setup. Everything lives in your repo, and the docs suggest checking agents into version control, so your team can read, change and reuse it.

Proof

Related work and reading

FAQ

Questions about this project

A sub-agent is a helper that Claude Code can hand a task to. Each one has its own context window, its own instructions and its own tool access, and it returns its result to the main chat. That keeps long searches out of your main chat and lets simple work run on cheaper models.

No. The Claude Code docs say CLAUDE.md files are context, not enforced configuration. For a rule that must never break, use permission rules or a hook that blocks the action. I use the rules file for guidance, and permissions and hooks for the hard limits.

Most likely, because Claude Code works with the code it can read. I know Laravel and PHP best, and I also work with JavaScript and TypeScript. Tell me your stack before you order, and I will say how I would set it up and which package fits.

Yes, on my own projects. I also wrote a comparison of Cursor and Claude Code. For Cursor I used its official docs and pricing pages. What I build for you is made for your repo, your rules and your budget, not copied from another setup.

Claude Code usage is billed to your own Claude plan or API account, so the price here is a one-time fee for the setup. Cost rules in the larger packages send simple jobs to cheaper models. I cannot predict your usage, because it depends on how much your team uses the agents.

Only if you trust the server. The Claude Code docs say to verify you trust each MCP server, because servers that fetch outside content can expose you to prompt injection. I connect only what you need, give each one the least access it needs, and explain what each one can reach.

Need something different?

If your work does not fit these packages, tell me what you need and I will reply with a plan.

More projects

From $500

Laravel MCP server

An MCP server for your Laravel app, so an AI agent can use only the tools you allow.

  • Token auth, scoped tools
  • Source code and setup guide
  • Works with clients like Claude

From $250

AI-built app audit

A senior engineer checks your AI-built app before you launch it.

  • Secrets, login, database, deploy
  • Ranked, plain-English report
  • Fixes in your repository

From $300

HubSpot integration

Contacts, companies and deals kept in step between Laravel and HubSpot.

  • Two-way sync with retries
  • Failed records logged and retried
  • Code in your repository

See all services