Thirteen weeks to an engineering team that builds with agents.
The Kelright Agentic Guild puts guardrails in your repository and CI that tell engineers and agents what to do as they work. A check in your systems closes every milestone, and part of our fee rides on the measured change.
Opens an email to hello@kelright.com.
The install, week by week
Deposit. When we sign.
M0 Diagnostic, weeks 0 to 2
M1 Instrumentation and cadence, weeks 1 to 3
M2 Test harness, weeks 3 to 6
M3 CI and delivery, weeks 5 to 9
Invoice. When M2 and M3 are accepted.
M4 Tech-debt tranche, weeks 7 to 11
M5 Agentic workflow adopted, weeks 8 to 12
Invoice. When M4 and M5 are accepted.
M6 Handoff, weeks 12 to 13
Outcome. At day ninety, part of the fee, on the measured change from your baseline.
How it engages
A fixed scope, a fixed fee, and a check that closes every milestone.
We start with the readiness check and a two-week diagnostic. Scope, fee and targets are written down before we begin, and they do not move.
- Who it is for
- Software teams of roughly eight to sixty engineers, writing code with agents or about to.
- Scope and fee
- Fixed before we start and set by team size, with each milestone's target in the statement of work.
- Your time
- A lead, two engineers who become champions, and a few hours of the team's time in the diagnostic.
- Access
- Read access to your repositories, CI and tracker. Every change we make is a pull request through your own review.
- Through a fund
- A fund can bring the Guild to its companies as a cohort. Each company decides for itself, and its numbers stay its own.
Who leads it
We lead your install, and we have shipped this way before.
- Agentic engineering before it had a name
- Head of engineering at a startup built with coding agents from the ground up.
- Five years forward-deployed at Palantir
- Landing inside a new company each time, and shipping in weeks. Founded its Pipelines Guild.
- A turnaround at a company more than fifty years old
- CTO of a global franchise business, through a full engineering and IT turnaround.
The milestones
Each milestone ends on a check, not a slide.
A milestone is done when your own systems say so. You never have to take our word for it.
M0: Diagnostic
Weeks 0 to 2
We read your repositories, pipeline and tracker and measure where you start: DORA, pull request cycle time, coverage, AI spend per merged pull request, and how fast a new engineer ships.
Accepted when the baseline is frozen in the Benchmark.
M1: Instrumentation and cadence
Weeks 1 to 3
The measures read live from your systems, and work runs on a weekly cadence in your tracker.
Accepted when every system you named reads into the Benchmark, and the cadence runs.
M2: Test harness
Weeks 3 to 6
Tests your agents and engineers can lean on, at an agreed coverage target, enforced in CI.
Accepted when CI coverage meets the target, and a change below it cannot merge.
M3: CI and delivery
Weeks 5 to 9
A pipeline fast enough to ship small changes often, with a safe way back.
Accepted when pipeline time and deploy frequency meet the agreed targets.
M4: Tech-debt tranche
Weeks 7 to 11
A debt register set at the start, cut into pieces one engineer or agent can take, and retired.
Accepted when every ticket in the register set at M1 is closed.
M5: Agentic workflow adopted
Weeks 8 to 12
Agent instructions carry your conventions, and agents take real work behind the gates, within a spend budget.
Accepted when agent pull request share, review time and AI spend per merged pull request meet the agreed targets.
M6: Handoff
Weeks 12 to 13
The playbook, two of your engineers trained as champions, and the software running without us.
Accepted when the guardrails run without us, and your champions change one through the gates.
How it is billed
We are paid as the checks pass, and partly on what changed.
A deposit when we sign, then two invoices, each after its milestones’ checks pass: M2 and M3, then M4 and M5. At day ninety, part of the fee is paid on the measured change from the baseline the Benchmark froze.
The fee is fixed by team size before we start; ask us for yours. After handoff, the software stays on as a subscription.
Where you start
The install starts from the phase your team is in.
The readiness check tells you your phase in about six minutes. The milestones are the same for every team; the targets are set from where yours starts.
The phases a company climbs
Fundamentals first
The basics are still going in: tests, a safe way to ship, numbers.
To climb: Tests you can lean on, and numbers everyone can see.
1. CheckThe readiness check finds your phase.A sound house, ready for agents
Sound foundations, and agents not yet at work on them.
To climb: Put agents on real work, behind the checks you already have.
2. BuildThe Guild climbs it with you.Agents in use, a couple of things behind
Agents do real work; one or two checks have not kept up.
To climb: Fix those before more of the work moves to agents.
Teams on that detour come back to this step: tighten the gates, then hand agents more.
3. MeasureThe Benchmark shows where you sit.Already compounding
Agents carry real work, checked before it merges.
To stay: Keep it as you grow, with a record of what you decided.
4. DecideThe Brain keeps the record.
What stays
The practice stays when we leave, because it lives in your repository.
- The guardrails
- Agent instructions with your conventions; CI gates that hold a coverage floor, require tests on agent changes and keep changes small; review rules to match.
- The Benchmark
- Still reading against the frozen baseline after we leave.
- The gap, as work
- The distance to the best teams on each measure, cut into one-person or one-agent issues and re-ranked weekly.
- Your champions
- Two engineers who run and extend the guardrails, with the playbook and an assistant that flags drift.