UCP for Food Ordering
UCP has a Food Technical Council but no food ordering spec yet. The council was seated on 16 July 2026 with Block (Square), DoorDash, Google, Toast and Uber Eats.
Its roadmap starts at checkout: identity linking, fulfillment, loyalty, tips, age-restricted items and order tracking come first. The menu catalog comes second.
Grocery-shaped pieces shipped in v2026-08-25 as shared capabilities: store locations, opening hours, pickup and items sold by weight. Restaurant ordering, with its modifiers, tips and delivery fees, is not expressible yet.
-
15 Jun 2026Charter proposed (#518)
-
16 Jul 2026Council seated
-
25 Aug 2026Grocery pieces ship in common/
-
Sep 2026Modifier proposals (#821, #824)
-
Not yetFirst food capability
01What the charter still has to specify
Phase 1 of the charter is checkout. This is how much of it UCP can express today.
| Area | Status | What it says |
|---|---|---|
| Modifiers and required choices | not yet | Food items carry size, add-ons, no-onions and required choices, so the menu and cart need deep, validating option trees. Two open proposals, nested options (#821) and selection rules (#824), aim at this; neither is a schema yet. |
| Live menus and availability | partly | For groceries, location lookup carries per-location item availability and opening hours. There is no menu or modifier schema, and no live status for items that sell out mid-session. |
| Age-restricted orders | not yet | The charter puts age checks (for example alcohol) in Phase 1 checkout. Agents need a contract for verifying eligibility before they commit an order. |
| Delivery windows, fees and tips | partly | Store hours and shipping-versus-pickup destinations became expressible in v2026-08-25. Service fees, delivery fees and tips did not. |
| Identity, loyalty and reorder | partly | Delegated sign-in and a loyalty extension shipped in common/. "Reorder the usual" and live delivery tracking are left to the personalization phase. |
02Background
Agentic food ordering now has a governing body. After Google named local food delivery as a UCP vertical, a Domain Tech Council charter for Food Ordering was proposed as UCP issue #518 on 15 June 2026, and the Food Technical Council was formally seated on 16 July 2026 with Block (Square), DoorDash, Google, Toast and Uber Eats as founding members. The charter laid out a three-phase roadmap.
Phase 1 is the checkout core (identity linking, a checkout that handles fulfillment, loyalty, tips and age-restricted items, and post-order tracking); Phase 2 adds the menu catalog, conversational cart and discounts; personalization comes last. There is still no published food-ordering schema, so this page stays the prescriptive read on what UCP for Food Ordering has to specify before agents can place a live order safely — and the charter's phasing tracks it closely. The v2026-08-25 release complicated that picture in a way worth reading carefully.
Food arrived in UCP — but through groceries, and through Shopping, not through the Food Ordering charter. The release calls out grocery readiness and ships location search and lookup, deterministic timezone-aware operating hours, fulfillment location context with explicit shipping-versus-pickup destinations, and fractional quantities with sale-basis pricing for items sold by weight — all as shared dev.ucp.common.* capabilities a grocer can declare today. Two charter Phase-1 building blocks also landed cross-vertically: delegated identity-provider sign-in, and a formalised loyalty extension.
What did not ship is a food namespace — the release carries only dev.ucp.shopping.* and dev.ucp.common.* — or the part that makes restaurant ordering hard: the modifier tree, live 86'd-item status, age-restricted checkout, tips and delivery-fee handling. The published roadmap scopes the vertical as restaurant and menu discovery across modification options, and dish selection through to checkout with tipping and delivery notes. Grocery pickup is expressible in UCP today; a Thai takeaway order is not.
What it would look likeSomeone tells an assistant "order my usual from the Thai place near me," and the agent places a live order — right item, right modifiers, valid delivery window, age check where it's needed — through the merchant's endpoint, with the restaurant staying the merchant of record.
03What to do now
Expose live availability
A menu an agent can trust means real-time item and store-status data, not a static catalog.
Model your modifiers cleanly
Required choices, option groups and add-ons need to be machine-readable and self-validating before an agent can build a correct cart.
Get discoverable now
Being agent-legible is the prerequisite, whichever vertical you're in.
04What we'll measure
05Food Technical Council (seated 16 Jul 2026)
06FAQ
Is there a UCP for Food Ordering spec yet?
No published schema, and no food namespace. The Food Technical Council was seated on 16 July 2026 and the charter (UCP issue #518) sets a three-phase roadmap, but the v2026-08-25 release ships capabilities under dev.ucp.shopping.* and dev.ucp.common.* only — nothing under a food namespace. There is no machine-readable food-ordering spec to validate against yet. We track it and will validate the moment it lands.
Who sits on the Food Technical Council?
The founding members announced on 16 July 2026 are Block (Square), DoorDash, Google, Toast and Uber Eats — the group that will author the food-ordering capabilities.
What does Phase 1 cover?
The checkout core: identity linking (unlocking loyalty status and rates, including new-user setup), a checkout capability covering fulfillment selection, loyalty, tips and age-restricted journeys like alcohol, and a get-order capability for tracking links and confirmation. The menu catalog, conversational cart and discounts come in Phase 2; detailed tracking, voice ordering and payment fingerprinting are Phase 3.
Why does the charter start at checkout instead of the menu?
Like the lodging charter, food ordering sequences the money-moving core first: identity, checkout and order confirmation are Phase 1, while menu discovery and cart building are Phase 2. The charter also leans on UCP primitives that already exist — identity linking and loyalty — rather than rebuilding them.
What namespace will UCP for Food Ordering use?
Not decided in public, and not published. As of the v2026-08-25 release the only namespaces in the spec are dev.ucp.shopping.* and dev.ucp.common.* — there is no dev.ucp.food.*. UCP's convention is a reverse-domain name whose authority is the governing body, and dev.ucp.* is reserved exclusively for capabilities governed by the UCP Technical Council responsible for that domain, so food capabilities will sit under dev.ucp.* once the Food TC publishes them. The exact segment is the Food TC's call and we won't guess it here — anything a vendor ships in the meantime belongs under its own domain (for example org.acme.menu), and our checker validates any declared namespace against the spec's authority-binding rule regardless of who defines it.
Did the v2026-08-25 release ship anything for food?
Yes — but for groceries, and through Shopping rather than the Food Ordering charter. The release ships location search and lookup, deterministic timezone-aware operating hours, fulfillment location context with explicit shipping-versus-pickup destinations, and fractional quantities with sale-basis pricing for goods sold by weight, all as shared dev.ucp.common.* capabilities. Delegated IdP sign-in and a loyalty extension landed too, which map onto the charter's Phase 1 identity work. Nothing restaurant-specific shipped: no menu or modifier schema, no 86'd-item status, no age-restricted checkout, no tips or delivery-fee handling. A grocer can declare UCP location and buy-by-weight capabilities today; a restaurant still cannot express its menu.
Is grocery the same vertical as food ordering?
Not in the spec. Google frames grocery as part of shopping and local retail fulfillment, and the v2026-08-25 primitives that serve it — location, hours, fractional quantities — shipped under dev.ucp.common.*, reusable by any vertical. Food Ordering remains a separate proposed Domain Tech Council charter (issue #518) aimed at restaurant and delivery flows, with no published namespace. In practice a supermarket is closer to a UCP retailer with locations; a restaurant needs primitives that do not exist yet.
What should restaurants and ordering platforms do now?
Get ahead of the predictable Phase 1 primitives: expose live item and store-status availability, model your modifiers and required choices as clean, self-validating option trees, and make sure fees and fulfillment windows are machine-readable — before you integrate.
07Keep reading
Other UCP verticals
UCP for Retail
LiveThe first and largest UCP vertical — agents search catalogs, build carts, and check out across thousands of verified stores.
19,019 verified merchantsUCP for Lodging
Draft capability merged · not yet releasedUCP's first lodging capability, dev.ucp.lodging.booking, is merged into the spec as a draft. It isn't in a release yet. Here's what it covers and what it still has to handle.
UCP for Travel
AnnouncedFlights, rail and travel distribution are agentic commerce's hardest vertical — perishable inventory and dynamically generated offers, not fixed SKUs.
UCP for B2B
Community-debated · no official UCP B2B spec yetB2B isn't an officially-named UCP vertical, but it's the most-debated one. Where UCP actually fits B2B today — and the primitives the spec would have to add for the rest.
UCP for Services
Community-proposed · no official UCP services spec yetAppointments, classes, reservations and rentals sell time, not stock. UCP has no services namespace yet — but a Governing Council working group is being explored, and scheduling capabilities are already turning up in real UCP profiles.
Explore protocol data
Is your ordering flow agent-ready?
Whichever vertical you're in, the prerequisite is the same: a valid, machine-readable UCP endpoint an agent can actually transact against.
Get the agentic commerce digest every Monday
Real adoption data, ecosystem trends, new spec versions, and the stores that broke or recovered this week. Read by founders and engineers building the next generation of commerce.
