AWAS (AI Web Action Standard) is an open-source specification that enables AI browsers and agents to interact with websites through machine-readable action definitions, without disrupting traditional human browsing.
Current AI browsers resort to slow browser automation that mimics human clicking. AWAS provides a structured, efficient way for AI agents to understand and execute website actions, similar to how Schema.org helps search engines understand content.
Yes! AWAS is designed to work alongside existing standards like HTML5, Schema.org, OpenAPI, and WAI-ARIA. It uses HTML data attributes and .well-known URIs that are ignored by traditional browsers.
- Website owners who want their sites to work with AI browsers
- E-commerce platforms looking to enable AI-assisted shopping
- SaaS applications wanting AI agent integration
- AI browser developers building agentic browsing experiences
- Basic (Level 1): 30 minutes - Just add the manifest file
- Standard (Level 2): 2-4 hours - Add HTML attributes
- Complete (Level 3): 4-8 hours - Full server integration
No! AWAS uses progressive enhancement. You add new attributes and files alongside your existing code without modifying it.
No. Traditional browsers completely ignore AWAS attributes and files. Your site works exactly the same for human users.
Yes! Start with one or two actions, test them, then gradually add more. You don't need to implement everything at once.
AWAS is language-agnostic. We provide reference implementations for:
- JavaScript (client-side)
- Python/Flask
- Node.js/Express
- PHP
- Ruby/Rails
Not necessarily. AWAS can work with your existing APIs. The manifest simply describes how to call them in a machine-readable way.
Yes, when implemented correctly. AWAS includes:
- Built-in rate limiting
- Authentication requirements
- Input validation specifications
- Action permission systems
- Audit logging capabilities
See SECURITY.md for detailed guidelines.
- Implement rate limiting (built into AWAS)
- Require authentication for sensitive actions
- Use action-level permissions
- Monitor AI agent activity through logs
- Restrict sensitive actions (checkout, payment) to human verification
No more than traditional bots. AWAS doesn't create new attack vectors—all the same security practices apply. In fact, AWAS makes it easier to identify and control AI agents.
We recommend requiring human verification for sensitive actions like checkout and payment. AWAS supports marking actions as requiring human confirmation.
- OpenAPI: Describes API endpoints and contracts
- AWAS: Bridges HTML interfaces with AI-executable actions
AWAS complements OpenAPI by adding web page context and UI element mapping.
- Schema.org: Describes what something is (entities, content)
- AWAS: Describes what you can do (actions, interactions)
They work together—use both for best results.
Yes! AWAS works with React, Vue, Angular, and other SPA frameworks. You can add attributes to your components and serve the manifest from your server.
Yes. For dynamically loaded content, AI agents can re-parse the DOM to find new actions, or you can provide API endpoints that return action definitions.
AWAS v1.0 focuses on HTTP-based actions. Real-time capabilities are planned for future versions. You can currently expose WebSocket endpoints in your manifest.
AWAS is a new standard. As AI browsers like Atlas and Comet mature, we expect them to adopt AWAS. Early adoption helps drive standardization.
While designed for AI browsers, the structured action format can benefit voice assistants, chatbots, and other AI agents.
Yes! AWAS works identically on desktop and mobile. The manifest and attributes are device-agnostic.
Yes! AWAS can be added to any CMS. We provide specific guides for popular platforms in the implementation docs.
- Better discoverability: AI browsers find and understand your site
- Reduced server load: Efficient API calls vs. DOM scraping
- Future-proof: Ready for the AI-driven web
- Competitive advantage: Early adopters rank higher in AI results
- Better analytics: Track AI agent interactions separately
Indirectly, yes. While traditional SEO focuses on search engines, AWAS helps with "AI SEO"—ranking in AI browser results and recommendations.
No! AWAS is completely free and open-source under the Apache 2.0 license. No fees, no royalties.
Absolutely! AWAS is community-driven. See CONTRIBUTING.md for guidelines.
Yes. AWAS includes privacy directives like:
ai-respect-privacy: trueai-no-data-training: true- Human verification for sensitive actions
AWAS doesn't change your privacy obligations. Apply the same privacy practices you use for traditional users. The audit logging features can help with compliance.
Only if you explicitly allow it. AWAS respects your existing authentication and authorization systems. Sensitive data should require authentication just like for human users.
Check:
- File is at
/.well-known/ai-actions.json - Server returns correct content-type:
application/json - File is valid JSON (use a validator)
- Discovery link is in HTML
<head>
Ensure:
- Middleware is properly installed
- AI agent detection is working (check for
X-AI-Agentheader) - Rate limit configuration is correct
- Redis or other storage backend is running (if used)
Verify:
- Action IDs match between manifest and HTML
- Endpoints are accessible
- CORS headers are set correctly
- Input validation passes
- Check server logs for errors
Add appropriate headers:
Access-Control-Allow-Origin: *
Access-Control-Allow-Methods: GET, POST, OPTIONS
Access-Control-Allow-Headers: Content-Type, X-AI-Agent
Roadmap includes:
- Workflow definitions (multi-step actions)
- Real-time update capabilities
- WebSocket support
- Integration with MCP and A2A protocols
- Visual editor for manifest creation
- Testing and validation tools
- Watch the GitHub repository
- Join Discussions
- Follow release notes and changelog
That's the goal!! Community adoption and feedback will help AWAS evolve into a formal web standard.
- Check the full documentation
- Ask in GitHub Discussions
- Open an issue
Last Updated: October 29, 2025