For teams building with AI agents

Govern your coding agents in a way you never could with people.

A person reads the spec once and remembers what they remember. An agent reads exactly what you hand it, on every run. PINKTURKEY keeps your decisions, requirements and constraints with a state on each — in force, or replaced — and hands the agent only what is in force today.

Create an account See how it works

Every decision carries its state: in force, or replaced
Replaced records never reach the agent
Ask what was in force on any date

The failure nobody is watching

A decision goes stale, and nothing tells the agent.

This is not a documentation problem. It is a control problem, and it grows with the speed of the agent rather than with the size of the team.

The decision ages
An agent writes an architecture decision. A month later the situation has moved on. The decision has not, and nothing marks it as past.
No state. No signal.
The agent keeps obeying it
The next run finds the file, reads it as current, and changes the implementation to match. The agent is doing exactly what it was asked to do.
Correct behaviour, retired input.
You find out late
By the time the code locks up, the complexity has passed what one person can hold, and you can no longer work out which document to remove.
The cost compounds quietly.

A product view, not a brochure diagram

One requirement, and everything currently holding it in place.

The record on the left is a requirement. Everything on the right is what currently constrains it: the decisions still in force, the interfaces that implement it, the evidence that still counts. This is the view the agent is given a scoped copy of.

What it saysThe accepted requirement and the version that is currently in force.
What holds itDecisions, interfaces, tests and evidence, each connected explicitly rather than inferred.
What the agent getsThe same picture, scoped to the task, with superseded records left out.

Connect your repository See what governs now

Payments modernizationGoverned workspace

Pilot workspace

REQUIREMENTS12 current

REQ-0042 · VERSION 8

Verify subscription activation before capability becomes available

ACCEPTED

Business capability becomes available only after trusted provider evidence is reconciled with the account and workspace activation state.

OwnerProduct & Billing

CriticalityHigh

ChangedToday

CONNECTED DELIVERY CONTEXT6 current links

DEC-0017constrainsProvider-confirmed activation

API-0012implemented bySubscription activation command

VRQ-0028verified byActivation state evidence

RSK-0009mitigatesPartial commercial activation

Compare plans

The mechanism

Every record knows whether it still governs.

The state is the product. Everything else follows from a record being able to answer one question: does this bind right now, for this scope, at this moment?

Authority state

A record is proposed, it governs, or it has been replaced. The third state is the one that does the work.

See what governs now

Scope and time

Authority is bound to a subject version, a scope and an effective period, so what governs can differ between two services and between two dates.

Typed relationships

Requirements, decisions and constraints are connected explicitly, so a change announces what it touches instead of leaving the model to infer it.

Governed context

The agent receives the records that are currently effective for its scope. Superseded records do not enter the context.

See the governed context

One door

A prompt, a bug report and a change request arrive the same way. The system works out which it is and relates it to what already exists.

Evidence on the record

What came back from a run is attached to the claims and versions it supports, and expires when they change.

See evidence and release

The economics

A folder of documents is an expensive way to tell a model what it needs to know.

Every run re-reads the same prose and re-infers the same relationships, and pays for both. Typed records with explicit relationships carry the same meaning in fewer tokens, and carry something a document cannot: whether the thing still applies.

Under test — not yet a measured result

The comparison is being run against a real repository rather than a benchmark.
The number will be published when it exists, including if it is smaller than we expect.
The claim already standing is correctness, not cost: a superseded record cannot reach the agent.

What is here today

The honest state of the product.

Some of this is in production. Some of it is not. Both are listed, because you would find out either way — and finding out here is cheaper for you than finding out three weeks in.

In production

Record identity, immutable versions, typed relationshipsAvailable
Authority state: what governs, what was supersededAvailable
Scope- and time-bound authority, as-of queriesAvailable
Policy evaluation with recorded decision evidenceAvailable

In build

Domain Hold — scoped blockingSchema is in production; the runtime path is the round in progress.Partial
Governed context: only effective records reach the agentIn build
One door for requests, changes and incidentsIn build
Cost estimate before the runIn build

Status is taken from the delivery plan, not from a marketing roadmap. An item moves to Available when it is in production and verified. The full list, with the same status column, is on the product page.

Per developer, monthly

Start on your own.

The individual plan is the whole product for one developer. The team plan is there when there is a team, and the enterprise route opens after the claims on this page have been measured rather than before.

Plan

Individual

For one developer who wants to know what is currently governing their codebase, and to stop retired decisions reaching the agent.

$29.00per month

Private engineering workspace
Governed specification records
Billing and invoice self-service
Platform-managed AI model route

AI usageEligible platform-managed route; limits and additional usage terms are disclosed before activation.

SupportStandard account, billing and operational support path.

AdoptionSelf-service account and monthly subscription.

Operating scopeIndividual professional

Start Individual

Recommended

Business

For a team that needs the same answer shared: one set of records in force, one governed context, one place requests and changes arrive.

$199.00per month

Up to 25 account members
Up to 3 governed workspaces
Workspace member and role management
Eligible AI model selection

AI usageEligible workspace model selection; provider and usage terms are disclosed before activation.

SupportStandard account, billing and operational support path.

AdoptionSelf-service account with governed team workspaces.

Operating scopeTeam / organization

Start Business

Plan

Enterprise

For organisations that need governed AI development across teams and workspaces, with evidence and release authority that can be audited.

$799.00per month

Up to 250 members by default
Up to 10 workspaces by default
Cross-workspace overview
Configurable governance limits
Controlled pilot onboarding and enterprise operating review
Bounded pilot onboarding and operating review

AI usageModel policy, provider boundaries and usage terms are defined through the pilot and commercial review.

SupportOperating expectations and support path are agreed during enterprise review.

AdoptionControlled pilot before broader estate rollout.

Operating scopeEnterprise estate

Talk to us about a pilot

Compare all plan capabilities

The boundary

It records which decision governs. You still decide.

The fastest way to lose an experienced reader is to imply the system knows what is correct. It does not, and it is not built to.

It does not decide what is correct

The system surfaces contradiction and tells you what currently binds. Judgement stays where accountability already is.

It does not trust a record it has not examined

An unexamined record cannot constrain an agent. What enters the system is checked before it is allowed to bind anything.

It does not run your pipeline

No deployment management, no ERP, nothing outside software development. The boundary is deliberate and it is not moving.

Start with one repository

The first run ends with a list of what currently binds in your code.

Connect a repository and PINKTURKEY reads what appears to be governing it today — decisions, constraints, requirements — and shows you which ones you had forgotten were still in force.

Connect your repository See how it works