[ AI tools · Development ]
Claude Code, With Senior Engineers in Charge
Our engineers use Anthropic’s Claude Code for migrations, refactors and test coverage across large codebases. We bring that way of working to your team, and we set it up so your own engineers can use it safely.
Plan4
- Migrate checkout to the new payments SDKReading the codebase
- Backfill tests for the billing moduleReading the codebase
- Upgrade the router and fix breaking routesReading the codebase
- Remove N+1 queries in the reports APIReading the codebase
Edit0
Test0
Ready for review0
Example board: four Claude Code sessions in parallel, each in its own git worktree. A senior engineer reviews every pull request.
Watch One Session
Claude Code is Anthropic’s agentic coding tool. It works on a whole repository and is strongest on changes too large or repetitive for a person to want to type. An example session:
Example session, shortened.
- 1Works where you workIn the terminal, VS Code and JetBrains, a desktop app, on the web and from Slack, on the same repository.
- 2Plans before it editsPlan mode reads the codebase and proposes the change; edits and commands run only within the permissions you set.
- 3Runs what it changesIt runs tests, linters and builds as it works and fixes what fails, so a pull request arrives with evidence.
- 4Many agents at onceSubagents for focused jobs and parallel sessions in separate worktrees for independent tasks.
- 5Extends with your toolsMCP servers, hooks, skills and plugins connect it to your tracker, docs and internal systems.
- 6Runs in CI tooA headless mode and GitHub Actions run the same flow for fixes, reviews and chores triggered from issues.
Every Repository Gets a CLAUDE.md First
Claude Code reads this file at the start of every session. It is the cheapest quality control there is, and the first thing we write on any project.
- 01
Commands it can trust
The exact commands to run the app, the tests and the linter. Without them, an agent guesses, and guesses waste a review cycle.
- 02
The team's conventions
Style and dependency rules, so changes look like the rest of the codebase and nobody has to rewrite them in review.
- 03
How quality is proven
Test-first bug fixes and rules about snapshots mean every pull request carries its own evidence.
- 04
The lines it must not cross
Hard limits, backed by permission settings and hooks that block the action, not just a polite request.
- 05
What the reviewer gets
A draft pull request with a summary, the test output and the agent's own list of doubts, ready for a senior engineer.
CLAUDE.md
Project rules for coding agents
## Commands
- - npm run dev, port 5199
- - npm test -- --run before every commit
- - npm run lint, fix all errors
## Code style
- - Match the surrounding code
- - No new dependencies without asking
- - TypeScript strict, no any
## Testing
- - Every bug fix gets a failing test first
- - Do not edit snapshot files by hand
## Never
- - Touch migrations already run in production
- - Commit secrets or .env files
- - Push to main
## Review
- - Open a draft PR with a summary and test output
- - List anything you were unsure about
Example file. Yours is written from your repository.
A Payments Migration in Six Reviewable Commits
Large changes go in small, reversible steps, each one reviewed and tested. The agent does the typing; engineers decide the order.
Six Rules We Never Relax
A coding agent makes a good team faster and a careless one dangerous. These are what keep it on the right side.
Permissions, not trust
Allowed commands and paths are set per repository; risky commands always ask first.
Hooks that enforce rules
Formatting, linting and secret scanning run automatically on every edit and commit.
Tests as the definition of done
No pull request without passing tests, and new tests for every change in behavior.
Senior review on every change
A named engineer reads every diff and owns what merges, whoever typed it.
Small pull requests
Changes are split so each one can be reviewed in under twenty minutes.
A record of the session
Plans and decisions are kept with the pull request, so the reasoning is reviewable too.
Rolling Claude Code Out Across a Team
What we configure when a company wants all its engineers on Claude Code, not just a few enthusiasts.
Plans and seats
Claude Team and Enterprise plans with central billing, or API usage billed per token.
Your cloud
Claude Code can run on models through Amazon Bedrock, Google Cloud or Microsoft Foundry, under your existing cloud agreement.
Managed settings
Organization-wide permissions, allowed tools and MCP servers pushed to every engineer's machine.
Single sign-on
SSO and role-based access on Enterprise plans, with usage analytics per team.
Data handling
Anthropic states that code and prompts under commercial terms are not used to train its models by default.
Cost control
Spend limits, usage dashboards and model choice per task, so routine work runs on faster, cheaper models.

[ Engineers who already work this way ]
Add an Engineer Who Ships With Coding Agents
A senior engineer joins your team within 48 hours, sets up CLAUDE.md and the guardrails in your repositories, and ships in your sprints.
Pick a Start, Then See the First Week
After you book
Book the Call- 01 · Today
A 30-minute call about your codebase and what you want done.
- 02 · Next day
A written plan: which work suits Claude Code, who does it, what it costs.
- 03 · Within 48 hours
An engineer starts, or the rollout kicks off with your first repository.
- 04 · Week 1
The first reviewed pull requests, with tests, in your repository.
Packages from $1,500; custom projects from $8,000.
What Engineering Leads Ask Us
>Is our code used to train Anthropic's models?
●Anthropic states that, under its commercial terms, code and prompts are not used to train its models by default. Teams that need their data to stay inside their own cloud agreement can run Claude Code on models through Amazon Bedrock, Google Cloud or Microsoft Foundry.
>Is AI-written code safe to merge?
●Only after the same review as any other code. Every change our engineers make with Claude Code is read by a senior engineer, covered by tests and scanned before it merges.
>What is Claude Code best at?
●Changes across many files: migrations, framework upgrades, refactors, test backfills and repetitive fixes. For a one-line edit, inline completion in your editor is usually quicker.
>Does it work with our editor and CI?
●Yes. It runs in the terminal, in VS Code and JetBrains IDEs, in a desktop app and on the web, and it has a headless mode and GitHub Actions for automated work in CI.
>Claude Code or OpenAI Codex?
●Both are strong coding agents, and our engineers use each. Claude Code tends to suit long, interactive work on a large codebase; Codex suits batches of well-scoped tasks in parallel cloud sandboxes. Many teams keep the same AGENTS.md and CLAUDE.md rules and use both.
>Can you set it up for our whole engineering team?
●Yes. Our team rollout covers repository rules, hooks, managed settings, MCP connections to your tools, CI workflows and working sessions with your engineers on real tickets.
Let's build your next product together
Book a free strategy call and leave with a clear plan and estimate. No commitment.
- 01Pick a time that suits you
- 0230 minutes on scope, stack, timeline and budget
- 03Fixed-price proposal, NDA on request