Claude Code Development & Team Setup
Software built with Claude Code, with an engineer reading every change.
Claude Code is an AI coding agent that works in a codebase: it reads files, runs commands and proposes changes. Used well, it compresses the slow parts of building software. Used carelessly, it produces changes nobody really read, in a codebase nobody fully understands.
We use it ourselves to build, and we set it up for teams that want the same speed with the safeguards in place. That means a project memory file that tells the agent how your codebase works, permissions that limit what it can run without asking, tests and CI that catch what a reviewer would miss, and a human code review before anything reaches production.
Two ways to work with us: we build the software with Claude Code and hand it over reviewed, or we configure Claude Code inside your repository and train your developers to use it safely. Either way the code is yours, in your repository.
What you end up with
Speed without unread code
Every change the agent writes is reviewed by an engineer, with tests and CI as the safety net. The agent drafts; a person decides what ships.
A codebase the agent understands
A project memory file (CLAUDE.md) that records your conventions, commands, architecture and the things the agent must never touch, so its work fits the code that is already there.
Permissions that match the risk
Allow-lists for safe commands, prompts for the rest, and hooks that run checks automatically, so the agent can move quickly inside limits you chose.
Your team able to carry on
Working practice written down and demonstrated on your own code: how to brief the agent, how to review its output, and when to stop and do it by hand.
What actually lands
Shaped to the engagement, but this is the shelf it comes off.
- 01Software built with Claude Code and reviewed by an engineer before handover
- 02Project memory (CLAUDE.md) and shared settings committed to your repository
- 03Permission rules, hooks and MCP server connections for your tools and data
- 04Test and CI set-up so agent-written changes are checked automatically
- 05Branch protection and a code-review routine for agent-authored changes
- 06A short team playbook: briefing, reviewing, and where not to use an agent
Where this has shipped
Claude Code, specifically
Yes. It is part of how we build. The agent drafts code quickly; an engineer reads and tests every change before it is merged, and the work is delivered in your repository with the tests and notes that let your own developers take it over.
Yes. That covers the project memory file, permission and hook settings, connections to your internal tools, and the review and CI routine around it, followed by a working session on your own code so the habits stick.
It is as safe as the controls around it. Keep it on a branch, limit what it can run unprompted, make tests and CI mandatory, and have a person review the diff. We would not point it at production data or credentials, and the setup is designed so it never needs to.
Yes. The school ERP case study is a system built with Claude Code and reviewed as pull requests, with seven read-only review agents and a CI workflow behind it. We have not published a rollout of Claude Code for a client's own developers.
No. We are an independent studio. We build on Claude through the public API and use Claude Code as customers do, and nothing here is endorsed by Anthropic.
Page copy drafted with AI assistance. Skofi is accountable for it. Editorial policy
Where this is applied
Tell us about the process that keeps costing you.
A 15-minute call is usually enough to say whether this is an automation, an integration, or something that should be built, and whether we are the right people for it.
What happens on the call
- 15 minutes, no deck
- We map the process on the call
- A written scope, or an honest no