design-partner phase

Ask. Approve. Done.

The AI DevOps engineer that manages your cloud, controlled via chat.

Become a design partner → See a live session

Limited to 3–5 partners · invite only

policy-service ok iam role scoped cloudtrail connected approvals required
01 — the problem

Infra work is interrupt work.

A bigger instance for the demo. A cert about to expire. Staging scaled down before the weekend and back up before standup. None of it is hard — all of it lands on the one engineer who knows the stack, usually mid-task, usually today.

So changes queue up behind a human bottleneck, get made by hand, and live in nobody's memory three weeks later.

02 — what is piper

An engineer, not a dashboard.

Piper connects to your AWS account through a scoped IAM role that you create — and can delete at any time. You describe what you need in chat. Piper reads the live state of your infrastructure, proposes a concrete plan, and waits. Nothing executes until you approve. No console to learn, no YAML to write: chat is the whole interface.

interface

Slack, today. Chat is the entire product surface.

access

One scoped IAM role — created by you, revocable by you.

approval

Required for every write. Reads are free; changes are gated.

audit

Native AWS CloudTrail. No separate log you have to trust.

03 — how it works

Three steps. One gate.

The approval gate is structural — not a setting you have to remember to turn on.

01

You ask in chat

Describe the change in plain language. Piper reads the live state of your infrastructure and drafts a concrete plan.

you › scale down staging for the weekend
02

Policy checks. You approve.

A separate policy service validates the plan against your rules. Then Piper waits — nothing runs until a human signs off.

policy pass · awaiting approval
03

Piper executes, fully logged

Piper applies the change with least-privilege credentials. Every API call lands in your AWS CloudTrail — not in a log of ours.

cloudtrail 14 events written

the approval gate is enforced by the policy service — piper cannot skip it

04 — live example

This is the whole interface.

A session from staging. Note what happens between “ask” and “done.”

P#infra-staging connected · us-east-1
james · fri 5:42 pm
Scale down the staging cluster for the weekend.
piper · fri 5:42 pm
Here’s the plan:
cluster
staging-eks
nodes
6 → 2
revert
mon 8:00 am, automatic
policy
✓ passed — production untouched
est. saving
~$118 over the weekend
Safe to run now. Approve?
✓ approved by @james · fri 5:43 pm
piper · fri 5:44 pm
Done — staging-eks is running 2 nodes. Reverting to 6 on Monday, 8:00 am. Every action is in your CloudTrail.
cloudtrail
14 events
revert
mon 8:00 · scheduled
the approve step is enforced by the policy service — piper cannot skip it
05 — who it's for

Small teams with real infrastructure.

Piper works best where infra toil is weekly and headcount for a platform team isn’t.

a good fit

  • 5–50 engineers, infrastructure on AWS
  • Staging and production you touch every week
  • Infra currently owned by “whoever is least busy”
  • Your team already lives in Slack

not a fit — yet

  • GCP or Azure-first — AWS comes first
  • Air-gapped or on-prem environments
  • Teams that want a fully autonomous agent — Piper won’t skip the approval gate
06 — design partner program

Build Piper with us.

3–5 spots only

We’re building Piper hands-on with 3–5 teams. In exchange for real workloads and honest feedback, you get an infra engineer that learns your stack — and a direct line to the people building it.

what you get

  • White-glove onboarding — we write the scoped IAM policy with you and review every permission.
  • A private Slack channel with the founding team. Same-day answers.
  • A weekly working session — your roadmap requests get built first.
  • Early access to everything we ship, before it ships.

what we ask for

  • A real AWS environment — staging first, production when you’re comfortable.
  • One 30-minute feedback call per week, for about ten weeks.
  • Honest, specific feedback — including where Piper falls short.
  • Tolerance for rough edges. This is early software, supervised accordingly.
security posture
Policy enforcement runs as a separate service.
IAM least-privilege access.
Full audit trail via native AWS CloudTrail.
Piper never touches production without your explicit approval.

Prefer email? Write to de.tanuson@mpmth.com and we’ll set up an intro.

07 — faq

Questions you should ask.

What access does Piper need?+
One IAM role in your AWS account, scoped to the services you want managed. We hand you the policy JSON; you review every permission before granting it. If you want to start colder, run the first week in read-only and watch what Piper would have done.
Can I revoke it instantly?+
Yes. Delete the role and access ends that second. There are no long-lived credentials on our side, nothing to rotate, no ticket to file. Typing /piper disconnect does the same thing from chat.
What happens if it makes a mistake?+
Piper only runs plans you approved, so there are no silent changes. If an approved change behaves badly, Piper rolls back to the prior state, and CloudTrail shows exactly what ran, when, with which credentials. During the partner phase, every incident gets a written postmortem from us.
What do design partners commit to?+
A real AWS environment (staging is a fine start), one 30-minute call a week, and blunt feedback. The program runs roughly ten weeks. No long-term contract — leave whenever, and delete the role on your way out.