Dopl connects AI agents to the devices people actually use, starting with smart glasses and growing to phones, computers, the smart home, cars and robots. Your agent sends one request that says what it wants to do, and Dopl decides which of the user's devices should receive it, checks that the user allowed it, holds the output to that device's limits and returns the result.
How it works
The cloud keeps device credentials in its own vault, so your agent only ever holds an opaque connection. Each device publishes the capabilities it supports, and every request is checked against what the user granted before an adapter carries it to the hardware.
What Dopl handles
- Connect. Your users pair a device from a hosted page that carries your brand, and they never need a Dopl account.
- Capabilities. One action model across every device, so your agent calls
present.displayorask.choicerather than a vendor API. - Permissions. Users grant capabilities by risk tier, and the most sensitive actions are denied unless confirmed, by rules that run outside any model.
- Context and routing. Each device kind has a profile that tells your agent how to write for it, which can be customized per project, and Dopl routes each request to the right device.
- Analytics. Every call records which device it reached, why, and how it went.
What works today
Agents reach Even Realities G2 glasses today through the hosted MCP server: they can show cards, ask questions with options and send notifications on the lens. The Device SDK, which adds project keys, a REST API, a TypeScript SDK, a project MCP server and the Connect flow, is live for design partners: Dopl issues your project's keys while the developer portal is being built, and @dopl/sdk is available on request. Every page in these docs carries a label that says whether it is live, in beta or coming soon.