v3 storefront nouns: media kit, playbook, business rules (preview)

Preview. /mcp/v3 is behind a feature flag and no accounts are enrolled. /mcp/v2/storefront is unchanged and nothing here deprecates it — see v3 Agent Surface for the shape of the surface as a whole. This page documents save_media_kit, save_playbook, and save_business_rules, plus what get and search return for them.

There is no merchandising noun on /mcp/v3. Merchandising is what your storefront does — the Merchandising Agent selling on your behalf — not a thing you save. Three nouns steer it, split by what kind of fact each one holds:

Noun	Kind	Owns
`media_kit`	material — what you show	Resolved identity (brand name, logo, membership badge), the authored story (summary, footprint, channels, regions, verticals, evidence links)
`playbook`	judgment — how you sell	Value-anchor pricing facts (floor/target/ceiling per condition), posture, packaging and selection guidance
`business_rules`	constraint — what you accept	The acceptance policy, and the two approval gates (`creativeApproval`, `mediaBuyApproval`)

This is a v3-only split. The four Pages you use today — Selling Terms, Approval settings, Acceptance Policy, and Merchandising Rules — are unchanged; see Merchandising for how those map to the platform. This page describes only the agent tools on /mcp/v3.

One fact, one owner

Each fact has exactly one noun that owns it, and this is enforced, not advisory: a playbook version whose content restates a price, a discount, a market, or an acceptance rule is rejected on save — save_playbook returns an error rather than a warning. The reason is drift: playbook is prose your agent reads, business_rules and media_kit are the structured facts downstream logic enforces, and if a $25 floor lived in both they would eventually disagree. Business rules is also where the platform’s enforcement lives, which is why it is a separate noun rather than a section of media_kit or playbook: the approval gate reads business_rules before anything else runs, and its absence fails closed (see Turning off human review below). media_kit and playbook are inputs your Merchandising Agent composes with; nothing gates on them directly.

Reading a noun: `get` takes no id

media_kit, playbook, and business_rules are singletons — exactly one per storefront — so get identifies them by kind alone. Passing an id for any of these three is a validation error.

{ "kind": "business_rules" }

{ "kind": "playbook", "include": ["versions"] }

include: ["versions"] adds the full version history for playbook and business_rules (an array of { version, isActive, notes, createdAt }); omit it and get returns only what is currently active. There is no separate “list versions” tool — version history is an include, the same way source diagnostics are.

`media_kit` — what you show

save_media_kit writes your authored story: a summary, channel and region footprint, verticals, and the evidence links it was sourced from. Omitted fields are left unchanged.

{
  "summary": "Regional lifestyle and news network reaching auto intenders across the Southeast US.",
  "propertyCount": 42,
  "channels": ["display", "ctv"],
  "regions": ["US"],
  "verticals": ["news", "lifestyle", "automotive"],
  "evidenceUrls": ["https://example-media-group.com/advertise/media-kit-2026.pdf"],
  "notes": "Extracted from the 2026 media kit; property count is an estimate."
}

{
  "mediaKit": {
    "identity": {
      "brandName": "Example Media Group",
      "logoUrl": "https://cdn.example.com/logos/example-media-group.png",
      "logoBackground": "dark-bg",
      "membershipStatus": "AAO_MEMBER"
    },
    "profile": {
      "summary": "Regional lifestyle and news network reaching auto intenders across the Southeast US.",
      "propertyCount": 42,
      "channels": ["display", "ctv"],
      "regions": ["US"],
      "verticals": ["news", "lifestyle", "automotive"],
      "evidenceUrls": ["https://example-media-group.com/advertise/media-kit-2026.pdf"],
      "notes": "Extracted from the 2026 media kit; property count is an estimate."
    }
  }
}

identity is read-only: brand name, logo, logo background, and membership badge are resolved from your registered brand identity, and save_media_kit has no fields for them.

What the kit can show today, and what it can’t yet. Brand name, logo, logo background, and membership status are the only resolved-identity fields a seller’s own read exposes. Your brand colours exist and are already shown to buyers and join-link visitors — but not back to you here, an asymmetry we know about rather than a design choice. Fonts, tagline, audience/reach figures, case studies, and tentpoles aren’t captured anywhere yet. There is also no price list of product templates on this noun yet — media_kit is its planned home, but no tool writes one today, so don’t expect save_media_kit to accept prices.

`playbook` — how you sell

save_playbook has two independent halves you can send together or apart: content (posture and packaging guidance, versioned) and pricing (your value-anchor facts).

{
  "content": "## Posture\nLead with premium CTV before offering display discounts. Hold value on flagship inventory during Q4.\n\n## Packaging\nBundle CTV pre-roll with display when a brief mentions video.",
  "notes": "Initial playbook, drafted from the Q4 sales deck.",
  "pricing": {
    "currency": "USD",
    "facts": [
      {
        "id": "ctv-flagship",
        "label": "CTV flagship pre-roll",
        "appliesWhen": "Buyer requests CTV pre-roll on flagship shows",
        "pricingModel": "cpm",
        "targetPrice": 42,
        "floorPrice": 32,
        "strength": "hard_floor",
        "provenance": "2026 rate card, page 4"
      }
    ]
  }
}

{
  "playbook": {
    "active": {
      "version": 3,
      "content": "## Posture\nLead with premium CTV before offering display discounts. Hold value on flagship inventory during Q4.\n\n## Packaging\nBundle CTV pre-roll with display when a brief mentions video.",
      "notes": "Initial playbook, drafted from the Q4 sales deck.",
      "createdAt": "2026-07-28T15:04:00Z",
      "ownershipIssues": []
    },
    "pricing": {
      "currency": "USD",
      "facts": [
        {
          "id": "ctv-flagship",
          "label": "CTV flagship pre-roll",
          "appliesWhen": "Buyer requests CTV pre-roll on flagship shows",
          "pricing": { "pricingModel": "cpm", "currency": "USD", "targetPrice": 42, "floorPrice": 32 },
          "strength": "hard_floor",
          "provenance": "2026 rate card, page 4"
        }
      ]
    }
  },
  "createdVersion": 3,
  "replacedVersion": 2
}

ownershipIssues names any line in the active version that restates a fact business_rules or media_kit already owns. pricing.facts replaces the whole list — a fact you don’t repeat in the request is removed, the same “whole document” semantics as content.

`business_rules` — what you accept

save_business_rules also has two independent halves: content (your acceptance policy, versioned) and the two approval gates.

{
  "content": "## Acceptance policy\nWe do not accept gambling, tobacco, or political advertisers. Fast-food advertisers require review before their creative goes live.",
  "notes": "Adds a review requirement for fast-food advertisers."
}

{
  "businessRules": {
    "creativeApproval": "manual",
    "mediaBuyApproval": "manual",
    "policy": {
      "version": 5,
      "content": "## Acceptance policy\nWe do not accept gambling, tobacco, or political advertisers. Fast-food advertisers require review before their creative goes live.",
      "notes": "Adds a review requirement for fast-food advertisers.",
      "createdAt": "2026-07-28T15:10:00Z"
    }
  },
  "createdVersion": 5,
  "replacedVersion": 4
}

Turning off human review

Moving creativeApproval or mediaBuyApproval to auto needs acknowledgeNoHumanReview: true on the same call:

{ "mediaBuyApproval": "auto", "acknowledgeNoHumanReview": true }

Leave it out and the save fails with a validation error rather than a silent no-op:

Moving media buys to `auto` means nobody reviews them, and your
acceptance policy is no longer consulted for media buys. Pass
acknowledgeNoHumanReview: true to proceed.

auto doesn’t relax your acceptance policy — it skips it. The gate that decides whether a media buy needs review is checked before the policy runs. With mediaBuyApproval: auto, every buy forwards and the active acceptance policy is never consulted at all — not “the policy allows it,” but “the policy doesn’t run.” With mediaBuyApproval: manual and no active policy, the opposite happens: nothing pre-clears and every buy is held for you. Activating a policy is what lets the clearly-on-policy subset forward on its own while you stay in manual.

Every save is create-and-activate — there’s no separate activate call

playbook.content and business_rules.content are versioned and immutable. Saving one creates a new version and makes it active in the same call — the response’s createdVersion and replacedVersion tell you what happened, and there is no activate input or activate_* tool on /mcp/v3. A version you save has full effect immediately; there’s no draft state where it exists but hasn’t taken effect. Rolling back is saving the earlier content again: doing so mints a new version with that content, which becomes active. There’s no “reactivate version 2” call — history is append-only and every version you can read is one you (or something writing on your behalf) actually saved. save_playbook and save_business_rules also each take an unversioned half in the same call — pricing on the playbook, the approval gates on business rules. When a call carries both halves, content is written and activated first. If the unversioned half then fails, the version is already live and is not rolled back — there is no transaction spanning the two. The error response still names createdVersion (and replacedVersion) so you know a version exists before retrying, and the retry only needs to repeat the half that failed: sending content again would mint a second, identical version rather than fixing anything.

Searching version history

media_kit, playbook, and business_rules are singletons, so search doesn’t search them — there is exactly one of each, and searching “your one playbook” isn’t a meaningful query. What is list-shaped is their version history:

{ "kind": "playbook_version", "query": "CTV" }

{
  "sources": ["objects"],
  "objects": {
    "kind": "playbook_version",
    "total": 3,
    "matched": 1,
    "returned": 1,
    "truncated": false,
    "results": [
      { "version": 3, "isActive": true, "notes": "Initial playbook, drafted from the Q4 sales deck.", "createdAt": "2026-07-28T15:04:00Z" }
    ]
  }
}

business_rules_version works the same way over acceptance-policy history. Omit kind and search lists every list-shaped object it knows about, playbook_version and business_rules_version included.

v3 Agent Surface

The one-endpoint, noun-shaped surface these tools live on.

Merchandising

The four Pages you use today, and how they map to the platform.

Acceptance policy

How the pre-screen reads your policy, and the three verdicts it returns.

Glossary

Component, Product, and the rest of the vocabulary these nouns build on.

Welcome

Core Concepts

Connect

Cross-Cutting Reference

Release Notes

v3 storefront nouns: media kit, playbook, business rules (preview)

One fact, one owner

Reading a noun: `get` takes no id

`media_kit` — what you show

`playbook` — how you sell

`business_rules` — what you accept

Turning off human review

Every save is create-and-activate — there’s no separate activate call

Searching version history

v3 Agent Surface

Merchandising

Acceptance policy

Glossary

​One fact, one owner

​Reading a noun: get takes no id

​media_kit — what you show

​playbook — how you sell

​business_rules — what you accept

​Turning off human review

​Every save is create-and-activate — there’s no separate activate call

​Searching version history

v3 Agent Surface

Merchandising

Acceptance policy

Glossary

One fact, one owner

Reading a noun: `get` takes no id

`media_kit` — what you show

`playbook` — how you sell

`business_rules` — what you accept

Turning off human review

Every save is create-and-activate — there’s no separate activate call

Searching version history