Prompt API

Concepts and usage

Using AI prompts to return information is very common on the web, with examples including rapid information look up, code and content generation, chatbots to provide customer service functions, image identification and description, audio transcription, and more.

The Prompt API provides an asynchronous (Promise-based) mechanism for a website to prompt the browser's own internal AI model. Having an on-device model is useful and efficient because sensitive data can stay on the user's device, the model is available offline, and developers can avoid the cost and latency of API calls to external services. The API abstracts away model-specific details such as tokenization and templating, so developers don't need to handle those differences across different implementations.

All interaction with the language model happens through a LanguageModel session. You can use this session to specify context for the model, such as providing documents, background information, or conversation history, and to prompt it for responses to specific questions.

Before creating a session, you can call the static LanguageModel.availability() method to determine whether the language model supports a given configuration on the current device. This allows pages to adapt gracefully if data for the desired configuration can't be provided or is not downloaded. For example, you could display a download prompt or fall back to a cloud-based AI service, rather than creating a session only to have it fail.

A session is created by calling the static create() method. Once you have a session, you can call append() to preload content into the session without generating a response, and prompt() or promptStreaming() to send text or multimodal input and receive the response.

You can cancel pending operations such as create(), prompt(), and append() using an AbortController.

After a LanguageModel instance has been created, you can release its assigned resources and stop any further activity by calling its LanguageModel.destroy() method. You are encouraged to do this after you've finished with the object as it can consume a lot of resources.

To get started, check out Using the Prompt API for a walkthrough of the basics.

Interfaces

CreateMonitor

Provides information on the progress of an AI model download, for example a language pack or some fine-tuning data.

LanguageModel

Represents a session with a browser-provided language model. Provides static methods for creating sessions and checking availability, and instance methods for prompting the model, managing context, cloning sessions, etc.

HTTP headers

Permissions-Policy; the language-model directive

Controls access to the prompt functionality. Where a policy specifically disallows its use, the LanguageModel.availability() static method will return unavailable, and any attempts to call other LanguageModel methods will fail with a NotAllowedError DOMException.

Security considerations

The Prompt API is restricted to secure contexts (HTTPS). In addition, creation of LanguageModel objects requires that the user has recently interacted with the page (transient user activation is required).

Access to the API is also controlled via the language-model Permissions-Policy directive.

Examples

For full examples, check out our guides, starting with Using the Prompt API.

Also see the Chrome dev rel team's demos:

• Prompt API playground
• MediaRecorder audio transcription
• Canvas API image prompt

Specifications

Standards positions

Two browser vendors oppose this specification. Known standards positions are as follows:

• Mozilla (Firefox): Negative
• Apple (WebKit): Negative

Browser compatibility

See also

• The Prompt API on developer.chrome.com (2026)