AgentReady

• 

  Logo : AgentReady — AI Commerce Readiness

• 

  Bannière 3:2 : AgentReady — Visibility. Understanding. Buyability.

Inspiration

AI agents are becoming a new way for people to interact with the web, but most small-business websites were built for humans and search engines — not for agents.

A business may have products, services, prices, booking options or contact flows publicly available, yet an AI agent can still struggle to reliably understand what the business offers or how to take action.

AgentReady started from a simple question: how can a small business know whether its website is actually ready for AI agents?

What it does

AgentReady audits a real website and evaluates its AI commerce readiness across three dimensions:

Visibility → Understanding → Buyability

It analyzes observable commercial information, evaluates what an AI agent can reliably understand, and produces evidence-based scores and prioritized recommendations.

The goal is not simply to make websites more machine-readable. The goal is to help businesses move from being visible to AI agents to becoming genuinely understandable and actionable by them.

AgentReady also exposes its audit capability directly to AI agents through WebMCP.

The native WebMCP tool agentready.run_audit allows an agent to launch the same website audit available to a human through the AgentReady interface and receive structured results including readiness scores, observed evidence, recommended actions, rendering provenance and monitoring eligibility.

This creates one capability for two users:

Humans get a clear visual audit. AI agents get a structured, callable tool.

How we built it

AgentReady uses an evidence-first audit architecture designed to avoid inventing commercial facts.

The challenge demo combines a browser interface with a bounded server-side audit function. For JavaScript-heavy websites where the initial HTML does not expose enough reliable evidence, AgentReady can use a controlled browser-rendering fallback and clearly reports that provenance.

The WebMCP integration registers agentready.run_audit through:

document.modelContext.registerTool(...)

The tool calls the same AgentReady audit capability used by the human interface and returns structured content that an AI agent can consume directly.

This means WebMCP is not a separate demonstration layer: humans and agents access the same underlying capability.

AgentReady also includes monitoring of comparable audits over time, allowing businesses to identify improvements or degradation in their AI readiness while avoiding false comparisons when an acquisition is incomplete.

Challenges

The hardest problem has been factual reliability across very different types of websites.

Real-world sites contain navigation elements, product cards, booking flows, duplicated prices, dynamic JavaScript content and commercial information distributed across different structures.

AgentReady therefore follows a simple principle:

Observed, not assumed.

If reliable evidence cannot be observed, the system should report uncertainty rather than manufacture a confident answer.

Another challenge was making dynamic website audits reproducible. The browser-rendering fallback and monitoring eligibility rules were designed to prevent incomplete page acquisition from creating misleading readiness changes.

What's next

The next step is expanding AgentReady from one callable audit capability into a broader AI commerce readiness layer where detected gaps can map to additional WebMCP actions.

The longer-term loop is:

Audit → Detect capability gap → Expose capability → Agent takes action

The vision is to make AI commerce readiness accessible to small businesses that do not have dedicated technical teams — so they can participate in an increasingly agent-driven web without having to understand the underlying technology.

Built With

• github
• html
• javascript
• netlify
• pytest
• python
• webmcp