Give every shopping agent the same retail truth.
The Dealer Agent Protocol gives AI agents one precise way to read dealer-published vehicles, verify availability and explain prices and offers, without exposing the DMS or CRM. It's open, it's free to implement, and every Dealer AI Website speaks it.
Dealer facts exist. Agent-ready meaning doesn't.
Inventory feeds, website pages, incentive programs and availability signals each describe a different part of the retail offer. An agent that scrapes and guesses can merge them incorrectly and quote a price or a vehicle that doesn't exist.
The protocol defines the answer contract: what a fact means, who published it, how fresh it is, and what remains conditional or unknown.
- Vehicle
- 2026 three-row SUV, GT-Line trim
- Availability
- Available · verified 2 min ago
- Advertised
- $74,590.00
- Required charges
- $899.00
- Conditional offers
- 2 · eligibility required
- Authority
- Dealer pricing + inventory sources
Five steps an agent follows instead of improvising.
Learn the gateway
Supported profiles, the dealer and its rooftops. get_manifest
Find published vehicles
Stable filters and cursors. inventory.search
Read the exact unit
Vehicle facts, media and source lineage. get_vehicle
Check what changes fast
Availability against an authority. verify_availability
Break down the price
Advertised price, required charges and conditional offers. get_disclosure
One contract. Three practical wins.
Control what an agent can claim.
Publish only approved retail facts and policies, including used-vehicle history, condition and inventory age. Internal records stay outside the boundary.
Used-vehicle truth →Build one agent-facing contract.
Map existing feeds and APIs behind a gateway and expose consistent schemas across rooftops without replacing systems of record.
Provider path →Stop reverse-engineering every dealer.
Typed, discoverable read tools for inventory and offers, with freshness, authority and uncertainty in every result.
Builder path →Retail facts in. Dealer controlled.
Version 0.1 focuses on the facts an agent needs to help someone understand a vehicle and its retail offer. MCP carries the calls; the protocol supplies the retail vocabulary, schemas, disclosure rules and conformance claims.
- Dealer and rooftop identity
- Published vehicle inventory
- Used mileage, history, condition and inventory age
- Authoritative availability
- Advertised price and dealer charges
- Discount and incentive conditions
- Provenance, freshness and uncertainty
Built in public for shared adoption.
Apache-2.0
The specification, schemas, examples, tests and reference gateway are open, with an explicit patent grant.
Read the license →Contribute
Propose protocol changes, improve schemas and tests, or strengthen the reference gateway.
GitHub →Created by our founder
The Dealer Agent Protocol and the reference gateway were created by Ariel Coro, founder of Dealer AI Websites, with contributions welcome from anyone.
Ariel Coro on GitHub →Every Dealer AI Website implements it.
You don't need to read a spec to benefit from it. Your site ships with a protocol-compliant gateway on day one, so any agent that speaks the protocol gets the same trustworthy answers from your store.