99% of websites aren't agent-ready: Ora benchmarks AI agents on live sites
yenkel · x · 2026-08-29
Ora has released a benchmarking platform built on Vercel that tests AI agents by executing real-world tasks like signups, integrations, and payments on live websites.
Key Findings:
- Tests reveal that 99% of the web is currently not ready for agents.
- The platform traces every step, showing where agents fail, what they tried, and what needs to change.
Technical Architecture:
- Front end, back end, and agent runtime all run on Vercel.
- Benchmarks every major agent side-by-side on actual customer sites.
Background: Ora's co-founder previously founded Tavily (an AI search engine acquired by Nebius), aiming to solve the next frontier: enabling agents to actually use the web products they find.
More from coding & agent
- Cursor to drop GPT model support starting Nov 12 — zainhas · 2026-08-29
- Geek Builds AI Trading Firm Using a Team of 10 Grok Bots — adamamcbride · 2026-08-29
- Engineer Tests Running an Agent Inside a Sandbox — jarrodwatts · 2026-08-29
- x402 Usage Doubles in Two Months: How to Access the Agentic Economy — 0xJeff · 2026-08-29
- AgentConnect launches as open-source, provider-neutral multi-agent alternative to Claude Tag — rohanpaul_ai · 2026-08-29
- Long Context Tests: Infinite Scroll Works Well in Personal Agents — vivekhaldar · 2026-08-29