We set up Claude Code across your repos in two days.

What we install

One configuration per repo, written into the repo so it travels with the code and survives a new laptop.

• Per-project departments

  Each repo gets its own set of agents and skills for the work that repo actually does. A website repo and a backend repo do not share one generic setup.

• Harness manifest

  A single file per repo that lists its agents, skills, cadences and model policy. Everything else is generated from it, so there is one place to change things.

• Quota guard

  Limits on how many runs, reviews and automated loops can fire per day, so one stuck loop cannot drain the account.

• Model-tier policy

  Cheap models by default for every routine task, including code. The top model is used only on an explicit signal: security, money paths, architecture decisions or a genuinely hard call.

• Safe-merge and review gate

  No direct pushes to main. Every branch is reviewed locally first, then by a second reviewer, and merged by a script that checks both passed.

• Memory conventions

  Where sessions write what they learned, how notes are indexed, and what never leaves the repo. The next session starts from the last one instead of from zero.

How it works

1. Step 1

   Before the install: a short intake

   You tell us which repos are in scope, who works on them and what already exists. We agree on the model-tier rules and the daily quota with you, in writing, before anything is installed.

2. Step 2

   Two days, remote, on your repos

   We install the setup repo by repo over a shared screen. Your team sees every file that lands and can ask about any of it. Nothing is merged without a person from your side approving.

3. Step 3

   Handover, then optional tune-up

   You get the manifest, the conventions and a walkthrough recording. If you want, we come back monthly to review what the fleet did, adjust the policy and clean up drift.

What you get

• A harness manifest and generated configuration committed in each repo in scope.
• A written model-tier policy and quota limits that your team can read and change.
• A working safe-merge gate on each repo, with the review steps it enforces documented.
• Memory conventions in place, with a worked example in each repo.
• A recorded walkthrough of the install and a short written handover.
• Optional monthly tune-up: a review of the past month, policy adjustments and cleanup of drift.

Price

Fixed prices. We confirm scope by email before any payment.

Start with a fleet health check

We audit your Claude Code setup against 12 known breakages and send you a written report. The audit is read-only: you share redacted config exports, we never touch your repos.

€390 Credited against the fleet setup if you go ahead.

Report in 3 working days.

Your email How many agent sessions do you run? (optional) I agree that Futurist Systems may email me about the health check. Privacy policy Join the waitlist

What this looks like in practice

What this is, and what it is not

• This is a setup of Claude Code, not a replacement for your engineers. The tool still needs people who read its output.
• We install conventions and guard rails. We do not promise a specific speed-up, a cost figure or a defect rate. We have not measured those across teams and will not quote numbers we do not have.
• The setup assumes your team already has Claude Code access and a git host. We do not buy or manage your subscriptions.
• Two days covers the install. Deep refactoring of your repos, CI migration or custom tooling is separate work and is quoted separately.

Frequently asked questions