# Agent Readiness

- **Event:** [Meta Global AI Developer Hackathon](https://cerebralvalley.ai/e/meta-ai-developer-hackathon)
- **When:** Oct 26, 12:00 PM – Nov 4, 9:00 PM (EST)
- **Where:** Online
- **Team:** [Vaibhav Deb](https://cerebralvalley.ai/u/Ezio_)
- **GitHub:** https://github.com/Ezio99/Hack_Agent_Readiness
- **Demo video:** https://www.youtube.com/watch?v=cgkxzWAKVDQ
- **Gallery:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery
- **Page:** https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1921

Agent Readiness — Making the Web Ready for AI Agents

The web was designed for humans, but increasingly, AI agents are navigating websites, gathering information, comparing options, and completing tasks on our behalf. Yet website owners have little visibility into whether their sites actually work for these agents.

Agent Readiness is an evaluation platform that measures how effectively AI agents can use a website, identifies where they struggle, and recommends actionable improvements.

Using Meta Muse Spark, the platform generates realistic tasks, executes autonomous browser agents in isolated environments, captures their actions and reasoning traces, and evaluates task outcomes. Website owners receive diagnostic reports, step-by-step replays, and measurable performance indicators.

To demonstrate its value, we evaluated two versions of the same educational website: an original version and an agent-ready version improved using the platform's recommendations.

In our canonical paired evaluation:
- Task success improved from 0/3 to 3/3.
- Model requests decreased from 39 to 10 (74% fewer).
- Execution time decreased from 326 to 57 seconds (83% faster).
- Model costs decreased from $0.571 to $0.134 (77% lower).

These results represent one controlled paired evaluation, not a claim of universal performance improvement.

The platform is deployed and publicly accessible, with real evaluation evidence, browser replays, diagnostics, and reproducible testing tools.

Live product: https://209-38-84-72.sslip.io

Public evaluation: https://209-38-84-72.sslip.io/#/reports/eval-1cf64d5eae90bcaeab531b4886140c14

Our longer-term vision is to establish agent readiness as a measurable property of websites, creating the foundation for an open registry that helps developers and AI agents understand which websites are ready for autonomous interaction.

## More from Meta Global AI Developer Hackathon

- [FirstLight](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1918)
- [Mind Over Model](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1919)
- [DryLine](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1920)
- [Desire-By-Design](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1922)
- [Lourdus](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1923)
- [SuryaR](https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1924)

---

Markdown version of https://cerebralvalley.ai/e/meta-ai-developer-hackathon/hackathon/gallery/1921. Site index for agents: https://cerebralvalley.ai/llms.txt · full text: https://cerebralvalley.ai/llms-full.txt