AI Agent Engineering Setup
For software teams already paying for Claude Code, Codex or Cursor whose agents still produce uneven work. A written Diagnostic shows what holds them back; a Setup Sprint installs repo contracts, shared skills, commit gates, one worktree per agent and spend tracking in your repos.
Step 1: Diagnostic, $750
- A read-only review of up to 3 repos: agent instructions (AGENTS.md, CLAUDE.md, Cursor rules), whether the build and test commands work on a fresh clone, and the CI gates agents must pass
- A baseline from your git history: merged pull requests per week, cycle time, the share of commits co-authored by an AI agent, and CI first-pass rate where CI data is shared
- A secret scan of tracked files, shared privately
- Your seat and spend picture, from a seat list you send (no keys)
- A written report with the top 10 fixes ranked, and a fixed quote for the Setup Sprint
Step 2: Setup Sprint, $4,500–$9,500, quoted in the report
- AGENTS.md and CLAUDE.md for each repo: architecture map, commands, conventions, never-do rules and a definition of done
- A shared skills library (fix a bug, add a feature, write tests, review a pull request, release, plus one for your stack) linked into each agent tool
- Commit gates: lint and tests before commit, a secret scan, and Claude Code hooks, with a documented, logged opt-out for people
- Parallel work: one git worktree and branch per agent, a task-slicing template, a review step, and a rule that people merge
- Seat, quota and spend tracking with alert thresholds, and a fallback plan for when a subscription runs out
- A before-and-after metrics report from your own git history, at team level, never ranking engineers
- A recorded walkthrough video of the workflow on your repo, and a written runbook your engineers can follow
- Async pull-request reviews on your engineers’ first agent-built changes, with written feedback
Not included
- Your Claude, OpenAI or Cursor subscriptions and API bills: your team keeps its own accounts and keys
- Merging to your default branch: agents open pull requests and your tech lead merges
- Production access or production data
- Vendor security questionnaires and certified-partner requirements
Delivery
Diagnostic in 5 business days from read-only repo access. The Setup Sprint takes 10 business days from your yes to the quote and covers up to 3 repos and up to 12 engineers.
You pay only the $750 Diagnostic here. The Sprint price depends on how many repos and engineers it covers, is set in the report, and starts only with your yes. The $750 is credited to it.
The Sprint is paid in full before it starts, like every offer here. It runs fully async, in writing: it needs about 30 minutes a day from your tech lead reviewing pull requests, and no meetings. Live calls are available at $100 per 15 minutes.
Everything installed is plain files committed to your repos, which you own. The scripts make no network calls, and the metrics read git history plus any CI and seat files you choose to share.
Simon runs App369 and Workspace369 this way: parallel Claude Code agents, one worktree each, shared skills, commit gates and a shared written memory.
Full refund if Simon can’t take it on once he sees your project, or if you cancel before work starts.
After you pay
Simon emails you within one business day to confirm and send the access checklist. You can also send it straight away to simon@app369.com. What the work needs:
- A read-only invite to up to 3 repositories (GitHub, GitLab or Bitbucket)
- Which agent tools your team uses (Claude Code, Codex, Cursor or others) and how many seats, with monthly spend if you know it. Never send keys
- Optional: a CI run export, and the two outcomes you most want from your agents
Delivery takes 5 business days from repo access. Everything happens in writing; no calls.
Full refund if Simon can’t take it on once he sees your project, or if you cancel before work starts.
Not sure it fits? See who it fits, or email a question about this offer; we reply by email, usually within one business day.