Home

AWAS (AI-readable Web Actions)

Introduction

AWAS (AI-readable Web Actions) is an open-source standard designed to enable AI browsers and autonomous agents to interact with websites efficiently and reliably. By providing machine-readable instructions about available web actions, AWAS bridges the gap between human-designed interfaces and AI-driven automation.

The standard allows website developers to explicitly declare how AI agents should navigate, interact with, and extract information from their web applications—ensuring more predictable, secure, and efficient AI-web interactions.

🎉 What's New in v1.1

AWAS v1.1 brings significant enhancements for predictable discovery, typed interactions, and safety contracts:

Enhanced Discovery

• HTTP Caching & ETags: Efficient manifest caching with ETag validators and 304 Not Modified responses
• Version Negotiation: HEAD request probing, AWAS-Version headers, and forward compatibility
• Predictable Updates: lastUpdated timestamp and specVersion fields for reliable version management

OpenAPI Interoperability

• Seamless Integration: Link AWAS actions to existing OpenAPI/Swagger specifications
• Reduced Duplication: Reference OpenAPI operations via documentUrl, operationId, or $ref
• Developer Friendly: Leverage familiar OpenAPI tooling while maintaining AWAS as the source of truth
• See OpenAPI Interoperability for details

Improved Safety & Typing

• Side Effect Levels: Explicit safe, idempotent, and destructive classifications
• Idempotency Support: idempotencyKeySupported and idempotencyKeyHeader fields
• Preconditions: Clear prerequisites for action execution
• Dry Run & Preview: dryRunSupported and previewUrl for confirm-before-write workflows
• Conformance Levels: L1 (read-only), L2 (write with preview), L3 (full transactional)

Better Type Safety

• JSON Schema: Full inputSchema and outputSchema support with examples
• Action Intents: Explicit read, write, delete, execute classifications
• Auth Scopes: Granular authScopes for permission control
• Rate Limit Hints: Per-action rate limiting guidance

Core Concepts

1. AI Manifest

A structured JSON file that describes available actions, endpoints, and interaction patterns for AI agents.

2. Action Definitions

Clear specifications of what actions can be performed, including parameters, expected inputs, and return values.

3. Navigation Guidance

Explicit instructions on how AI agents should navigate through the website to accomplish specific tasks.

4. Data Schema

Standardized formats for data exchange between AI agents and web applications.

5. Security & Authentication

Guidelines for secure AI-web interactions, including authentication methods and access control.

6. Rate Limiting & Ethics

Best practices for responsible AI behavior, including rate limits and usage policies.

Wiki Index

Explore the following pages to learn more about AWAS:

Specification

Complete technical specification of the AWAS standard, including syntax, structure, and validation rules.

Implementation Guide

Step-by-step guide for implementing AWAS on your website, with code examples and best practices.

Security Practices

Security considerations, authentication patterns, and guidelines for protecting your application from malicious AI agents.

FAQ

Frequently asked questions about AWAS, covering common use cases and troubleshooting.

Example AI Manifest

Real-world examples of AWAS manifests for different types of applications.

Developer Resources

Tools, libraries, validators, and community resources for working with AWAS.

---

Getting Started

To get started with AWAS:

1. Read the Specification to understand the standard
2. Review the Example AI Manifest to see practical implementations
3. Follow the Implementation Guide to add AWAS to your website
4. Consult Security Practices to ensure safe deployment

Contributing

AWAS is an open-source project. Contributions, feedback, and suggestions are welcome on the main repository.

License

This project is open source. Please refer to the repository for license details.