Ten working days.
Your codebase is agent-ready.
Most teams have Claude or Copilot switched on and are still getting slop out of it. That is not a model problem. It is that nobody told the agent how this codebase works. You send me the repo. I fix that async and return the artifacts within ten working days. You keep everything I build.
Your agents are guessing
An agent pointed at an undocumented codebase does what a new hire would do without onboarding: it invents conventions, ignores your invariants, and writes code your reviewers send back.
The fix is not a better prompt. It is the same thing that makes human onboarding work: written conventions, real tooling access, and clear limits on what can be touched.
That setup takes a specialist about ten working days of focused work. Most teams never get around to it, so they conclude the tools do not work.
Fixed deliverables
Everything below is built against your repository, async. No scheduled team-days, no calendar Tetris.
Repo audit
Where agents will and will not work in this codebase, and exactly why.
Written conventions
A CLAUDE.md for each repo in scope, tuned to how your team actually writes code, not to a template.
A custom MCP server
Wired to one internal system your team uses every day: your ticket tracker, your staging database, your internal docs, your design system. Yours to keep and extend.
Three working agent workflows
Built live against your real backlog. Not demos.
A code-review skill
Your team's actual review standards, encoded so the agent applies them before a human sees the PR.
Guardrails
What agents may never touch, how secrets are handled, how changes get gated before merge.
A 30-day adoption plan
One page, plus a checkpoint call at day 30.
Loom walkthrough on handoff
One Loom video walking through every deliverable for anyone who joins later.
Format: fully async, delivered against your real repo. One 25-minute kickoff call to align, then written check-ins as I ship.
The ten days, milestone by milestone
You may already be paying for this
Under the Loi favorisant le développement et la reconnaissance des compétences de la main-d'œuvre (the “loi du 1%”), Quebec employers with a payroll above $2M must spend at least 1% of payroll on eligible training. Whatever you do not spend, you pay to the Fonds de développement et de reconnaissance des compétences de la main-d'œuvre instead.
The rules were loosened in April 2026, and more training now qualifies than most employers realise.
If you are over the threshold, the budget for this already exists. I am happy to walk your finance contact through it.
Eligibility depends on your specific situation. Confirm with your finance team or Revenu Québec.
Four rungs, one path
Agent Readiness Review
25 minutes on a call, plus a one-page written finding: where your codebase blocks agents, what it is costing you, and the three fixes worth doing first. No obligation, and useful whether or not you book anything.
Executive session
A 60-minute call with your CTO, VP Engineering, and team leads, plus a written brief. What agents change about how software gets built, where the real risk sits, and what to fund next quarter.
Groundwork
Ten working days async, all eight deliverables above, against your real code. One 25-minute kickoff call, written check-ins as I ship, Loom walkthrough on handoff.
30-day support
Async access plus a checkpoint call while the team builds the habit.
Prices in CAD. Delivered in English or French.
What I bring to the room
I have shipped 42 projects across AI, data, healthcare, hardware, and developer tooling, and won four hackathons doing it. I have built production MCP servers, multi-agent orchestration, and agent workflows against real systems, and I know precisely where they break, because I have broken them.
Day to day I am a senior developer working on enterprise integrations. Groundwork is what I have learned about making these tools survive contact with a real codebase.
Start with the free review
Twenty-five minutes. You get a written one-pager either way.
Book the 25-minute readiness review