Aller au contenu principal

About DoorDash MCP

Private beta

DoorDash MCP is in private beta. Access is currently limited to approved testers.

Want in? Join the waitlist and we'll reach out regarding next steps.

DoorDash MCP is DoorDash's Model Context Protocol server. It lets an AI agent search, build a cart, and order from DoorDash on behalf of a signed-in consumer, using the same catalog, cart, and checkout systems that power the DoorDash app.

DoorDash MCP is built for corporate and organizational ordering: agent platforms and internal tools that allow employees of a business to place DoorDash orders. For example, an internal assistant that orders lunch for a team, or a corporate travel/ops tool that expenses meals. If you're building a personal, single-user workflow instead, see DoorDash CLI.

DoorDash MCP is not currently available to be integrated into consumer-facing products.

What an agent can do​

Once your MCP client is connected and a user has authorized access, your agent can:

  • Discover — find restaurants and retailers near an address, browse menus, and look up items
  • Build a cart — add and remove items, apply promotions, and preview pricing and totals
  • Manage delivery details — read and set a consumer's saved delivery addresses and default payment method
  • Order — submit an order, check its status, fetch a receipt, and reorder from order history

See Available tools for the full list of tools exposed to the corporate-ordering tier, with parameters and return values.

How it works​

  1. Your MCP client connects to DoorDash's MCP endpoint.
  2. The consumer authorizes your client through DoorDash's OAuth consent flow — the same "Sign in with DoorDash" flow used elsewhere in the DoorDash ecosystem. Your agent never sees or handles the consumer's DoorDash credentials directly.
  3. DoorDash issues your client an access token scoped to that consumer.
  4. Your agent calls MCP tools using that token.

Get started​

Ready to build? Head to the quickstart to connect your first MCP client.