Food Delivery Agent Readiness: Why DoorDash Has APIs But Restaurants Don't
DoorDash, UberEats, and Grubhub have merchant APIs that let software manage menus, orders, and delivery. The restaurants on those platforms have zero. The middleman is agent-ready. The actual business is not. This is the biggest disintermediation opportunity in the agent economy.
The Platform Has the API. The Restaurant Does Not.
There are over 1 million restaurants in the United States. Roughly 350,000 are listed on DoorDash. Another 900,000 are on UberEats. All three major delivery platforms offer merchant APIs: DoorDash has its Drive API and Storefront API, UberEats has its Orders API with menu sync and delivery webhooks, and Grubhub provides a merchant portal with limited programmatic access.
But here is the critical asymmetry: the platforms are agent-ready, the restaurants are not. When an AI agent needs to order dinner for a user, it has two options. It can call the DoorDash API and place an order through the platform, paying a 15-30% commission that the restaurant absorbs. Or it can try to interact with the restaurant directly, and find nothing — no API, no structured menu, no ordering endpoint. Just a phone number and maybe a PDF menu.
The agent takes the path that works. Today, that path goes through the platform. Every time.
Platform vs Restaurant: Agent Readiness Comparison
Delivery platforms invested in APIs because they are technology companies. Restaurants did not because they are food businesses. The result is a massive agent readiness gap between the middleman and the actual service provider.
The 50-point gap: DoorDash scores an estimated 52 for agent readiness. The average restaurant on DoorDash scores 8. That is a 44-point gap between the platform and the business it represents. The platform captures all agent traffic and charges 30% for the privilege.
The 30% Commission Trap
Restaurant profit margins average 3-9%. Delivery platforms charge 15-30% per order. The math is brutal: a restaurant making 6% profit on a $30 order earns $1.80. DoorDash takes $6-9 from that same order. The platform makes 3-5x what the restaurant makes on every delivery order.
Restaurants accepted this trade because they had no alternative. They could not build their own ordering technology, could not handle delivery logistics, and could not reach customers who had already shifted to app-based ordering. The platform owned the customer relationship and the technology layer.
AI agents change this equation. An agent does not open the DoorDash app. It queries available ordering endpoints. If the restaurant has its own MCP server with a create_order() tool, the agent calls it directly. The platform is bypassed entirely for the ordering step. The restaurant keeps 100% of the order value and pays only for delivery logistics if needed.
What an Agent-Ready Restaurant Looks Like
An agent-ready restaurant exposes its own ordering capabilities through an MCP server. Five tools cover the entire customer interaction — from browsing the menu to placing an order.
get_menu()Full menu with prices, descriptions, dietary tags, photos, and availability. Structured JSON, not a PDF.
check_availability()Real-time table availability, delivery radius, estimated wait times. No more calling to ask.
create_order()Place a pickup or delivery order with itemized items, modifications, and payment token. End-to-end.
get_delivery_estimate()Delivery time and fee based on address. The restaurant controls the price, not a platform.
get_specials()Daily specials, happy hour deals, catering packages. Agents find them instantly.
With these five tools, an AI agent can handle the complete ordering flow. A user says “order me a large pepperoni pizza from Mario's for pickup at 7pm” and the agent calls get_menu(), finds the item, calls check_availability() to confirm the time slot, and places the order via create_order(). No app download. No platform commission. No phone call.
The Disintermediation Play: 30% to 0%
This is the same pattern that played out with hotels and OTAs. Booking.com and Expedia built agent-accessible booking systems and charged hotels 15-25% commission. Hotels that invested in direct booking technology clawed back margin over time. The hotel industry spent a decade and billions of dollars on “book direct” campaigns.
Restaurants have the same opportunity right now, but at a fraction of the cost. An MCP server from AgentHermes takes 5 minutes to set up and costs less per month than a single DoorDash order's commission. Every agent-driven order that comes through the MCP server instead of DoorDash is pure margin recapture.
The math on a $30 average order: DoorDash takes $6-9. Direct MCP ordering costs the restaurant under $0.10 in infrastructure. That is a 98% reduction in ordering costs for every order that shifts to the direct channel.
Phase 1: Coexist
Keep DoorDash for human app users. Add MCP server for agent-driven orders. Two channels, additive revenue.
Phase 2: Shift
As agent traffic grows, more orders bypass the platform. Restaurant captures growing share at 0% commission.
Phase 3: Own
Restaurant owns the ordering relationship. Platform becomes delivery-only service at flat fee, not percentage.
First-mover advantage: The first restaurant in each neighborhood with an MCP server gets every agent-driven order for that cuisine category. AI agents prefer structured, direct endpoints over platform intermediaries. Being first is not just an advantage — it is a monopoly on agent traffic for your category and location.
Why Restaurants Have Not Built This Themselves
No technical team
Restaurants are food businesses. They have chefs, servers, and managers — not developers. Building an API is not a weekend project for a restaurant owner.
POS fragmentation
Toast, Square, Clover, Aloha, Revel — the POS market is fragmented across dozens of systems. Each handles menus and orders differently. No universal standard.
Delivery dependency
Platforms bundled ordering with delivery. Restaurants think leaving DoorDash means losing delivery capability. It does not — delivery-as-a-service exists separately.
Nobody told them
Restaurants do not know what an MCP server is, just like they did not know what a website was in 1996. The awareness gap is the real blocker, not the technology.
Every one of these barriers is solvable with a platform approach. AgentHermes auto-generates MCP servers with restaurant-specific tools pre-configured. The restaurant owner fills in their menu data, business hours, and delivery area. The MCP server handles the rest. No coding, no servers, no POS integration required for the first version. Connect POS later for real-time menu sync — but start capturing agent orders now.
Frequently Asked Questions
Why do delivery platforms have APIs but restaurants do not?
Delivery platforms are technology companies that happen to deliver food. They built APIs to onboard merchants, sync menus, and manage orders programmatically. Restaurants are food businesses that happen to need technology. Building an API was never on their roadmap, just like building a website was not in the 1990s. That gap is the opportunity.
Can a restaurant really bypass DoorDash with an MCP server?
For agent-driven orders, yes. When a user asks an AI assistant to order dinner, the agent looks for structured ordering endpoints. If the restaurant has an MCP server with create_order(), the agent calls it directly. No DoorDash, no commission, no markup. The restaurant keeps the full margin. This does not replace delivery logistics — it replaces the ordering middleman.
What about delivery logistics if the restaurant goes direct?
Restaurants can use delivery-as-a-service providers like DoorDash Drive, Uber Direct, or local couriers. These charge a flat fee per delivery ($5-8) instead of 30% of the order total. The restaurant handles ordering via MCP, and outsources delivery at a fraction of the platform commission.
How hard is it for a restaurant to get an MCP server?
With AgentHermes, it takes under 5 minutes. The restaurant connects through our wizard, selects the restaurant vertical, and we auto-generate an MCP server with menu, availability, ordering, and delivery tools pre-configured for the food service industry. No coding, no servers to manage.
Skip the 30% commission
See your restaurant's Agent Readiness Score, then get an MCP server that lets AI agents order directly. No platform fees, no app downloads, no phone calls.