Skip to content

Prebuilt mandates

Nine outcomes. Criteria included, priced before you ask.

Each one has shipped before: a purpose-built agent, a fixed price, and acceptance criteria you can read right here. Open any card. No estimate round, no review cycle; a prebuilt mandate takes an open slot and starts.

For
PM-01Delivered 12×

Coverage to a stated bar

Test coverage raised to the number you set on the ingest and dispatch paths.

From 18,000 · fixed

Acceptance criteria, as signed

  • Coverage on the named paths at or above the signed bar (met)
  • Every new test passes 30 consecutive CI runs (met)
  • No production code changed except test seams, reviewed (met)
  • Coverage report wired into CI, visible per pull request (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe
PM-02Delivered 10×

CVE burn-down to zero criticals

Advisories resolved, pinned, proven closed across carrier integrations.

From 16,000 · fixed

Acceptance criteria, as signed

  • Zero critical and high advisories open at acceptance (met)
  • Every dependency pinned with a stated upgrade path (met)
  • CI gate blocks reintroduction of known CVEs (met)
  • Change log signed per fix, no behavior regressions (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe
PM-03Delivered 7×

Flaky test eradication

Intermittent failures fixed and held green 30 days in the scheduling suite.

From 14,000 · fixed

Acceptance criteria, as signed

  • Named flaky tests pass 200 consecutive runs (met)
  • Root cause documented per test, not retries added (met)
  • Suite runtime does not grow more than 5 percent (met)
  • 30-day green streak monitored and reported (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe
PM-04Delivered 3×

Framework major upgrade

One major version jump behind flags, zero regression on the TMS core.

From 22,000 · fixed

Acceptance criteria, as signed

  • Target version live behind a reversible flag (met)
  • Full regression suite green on old and new paths (met)
  • Performance within 3 percent of baseline, measured (met)
  • Rollback rehearsed and documented in the runbook (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe
PM-05Delivered 8×

Observability baseline

Traces, dashboards and alerts, documented across yard and dock telemetry.

From 15,000 · fixed

Acceptance criteria, as signed

  • Golden paths traced end to end with named spans (met)
  • One dashboard per service your on-call can read (met)
  • Alerts page on symptoms, not causes, with runbooks (met)
  • A verified incident drill run with your team (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe
PM-06Delivered 6×

ETL pipeline hardening

Idempotent, monitored, replayable ingest on shipment event ingest.

From 19,000 · fixed

Acceptance criteria, as signed

  • Every job idempotent, proven by double-run tests (met)
  • Failed batches replay from checkpoint, no manual steps (met)
  • Data contract published and enforced at the boundary (met)
  • Freshness and volume alerts with stated thresholds (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe
PM-07Delivered 5×

Dead code and dependency purge

Unused paths removed with proof nothing broke in the dispatch console.

From 12,000 · fixed

Acceptance criteria, as signed

  • Unreferenced code removed, listed in the evidence pack (met)
  • Bundle or build size reduced by the signed target (met)
  • Full suite green, error budget untouched for 14 days (met)
  • Dependency count reduced with licenses re-audited (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe
PM-08Delivered 4×

API contract and versioning

Public surface documented, versioned, enforced on the rating engine.

From 17,000 · fixed

Acceptance criteria, as signed

  • OpenAPI spec generated from code, drift-checked in CI (met)
  • Breaking-change gate blocks unversioned changes (met)
  • Deprecation policy published with sunset dates (met)
  • Consumer smoke tests run against every release (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe
PM-09Delivered 6×

CI to 10 minutes

Pipeline cut to a signed wall-clock bar across partner EDI feeds.

From 16,000 · fixed

Acceptance criteria, as signed

  • Median pipeline under the signed bar for 30 days (met)
  • Flake rate under 1 percent across the window (met)
  • Cache and parallelism documented in the runbook (met)
  • No test coverage removed to hit the number (met)

Warranty 90 days · evidence pack on acceptance

Start via the probe

Every mandate below has shipped for a logistics operator. The examples name their systems, not ours.

New to the model? Read what a mandate is

Standing mandates

The audit comes every year. The CVEs come every week.

Someone in your company answers the insurance questionnaire and faces the auditor, and the evidence they need is engineering work nobody scheduled. A standing mandate fixes that: one named outcome delivered every month with proof, at a printed monthly price. That person is the buyer. Engineering is consulted, never sold.

Custody means the CVE queue is ours to hold: we watch it, we patch it, and we sign the proof. Your engineers stay on the product.

  • Starts from a clean baseline: the CVE burn-down (PM-02) runs first, and its price is credited as month one.
  • Prepaid quarterly. Cancel any month. Works on any codebase, not only ours.
  • The last page of every pack names the one thing we would fix next, priced. One item, never a list.

The probe report says whether custody fits your codebase, before any money moves.

One month of custodySample data
NEW ADVISORIES · WEEKLYIN OUR CUSTODYwatched · patched · provenCRITICALS OPEN0at month closeCVE-2026-30154criticalCVE-2026-29881highCVE-2026-31207criticalCVE-2026-30154patched · PR + permanent testCVE-2026-29881patched · PR + permanent testCVE-2026-31207patched · PR + permanent testEVIDENCE PACKauditor + insurer readableSIGNED · MONTHLY
PM-10Monthly
Cancel any month

CVE custody

New criticals fixed inside 7 days, zero at month close, with evidence.

Signed every month

  • Zero critical advisories open at month close; new criticals closed inside 7 days
  • High advisories resolved inside 30 days, each with a stated upgrade path
  • Every fix ships as a reviewed pull request with a permanent test
  • Monthly evidence pack your auditor and your insurer can read, signed by an engineer, with the supply-chain affected-or-clean attestation during an ecosystem event

Base

2,500 / month

Up to 5 repositories

Evidence

4,500 / month

Multi-repo, with the PCI and SOC 2 pack

The probe report says whether custody fits your codebase, before any money moves. Run the free probe

The full page →

The same custody, for the apps your team built

AI made building the app easy. Keeping it alive is nobody's job.

Someone on your team shipped a real internal app with Claude Code. Now the Monday meeting runs on it, and its builder became the support desk. App custody makes keeping it alive our job: a named owner, a timed restore every month, and evidence that it works.

  • Bought by the operating or technology leader who now owns the app estate: the ops, RevOps, or finance lead who built it, or the IT director, COO, CTO, or chief of staff who inherited it.
  • Starts from the free census: one page per app and the count. Exclusions are named at onboarding: no new features, no redesigns, no migration off the current stack.
  • Prepaid quarterly. Cancel any month. Repository-backed Claude Code projects are supported; other build environments are confirmed before any money moves.
PM-11Monthly
Cancel any month

App custody

The apps your own people built, kept alive and provably so, every month.

Signed every month

  • Every covered app has an owner of record, swept dependencies and secrets, and reviewed access
  • A restore is actually performed and timed every month, and a smoke test covers the paths the team really uses
  • One small fix per app ships as a reviewed change, signed by an engineer
  • The register stays current: one page per app, written for whoever asks what you run

Watch

1,500 / month

Up to 3 apps

Base

2,500 / month

Up to 8 apps

Evidence

4,500 / month

Up to 20 apps, or any app touching customer or payment data, with the register written for an auditor

The census is the door: one 30-minute interview produces the register, one page per app, and the count. Free. Book the free census

The full page →

Where we say no

Work we turn down. A fixed price only works if we say no a lot.

No

Scope we cannot measure

If we cannot write criteria you would sign, it is not a mandate yet. It becomes a paid diagnostic until it is.

No

Hourly, seats, or retainers

Buying our time makes the risk yours again. We would rather lose the deal than sell the wrong unit.

No

Work below the floor

Under 12,000 the review time costs more than the outcome earns. Small asks belong in a prebuilt mandate.

No

More mandates than the gate can sign

Agents scale; accountability does not. Every mandate ships under a senior who actually read it. We take what the gate can sign, and give the rest a date.

Not seeing your outcome?

Bespoke mandates start the same way: the probe reads your codebase, Cambrian sizes the work into criteria and prices, and a senior engineer signs the estimate. If we cannot write criteria you would sign, we say so.

Run the free probe