Observed arrival · 2026-10-16
agentscanr: AI agents put through UK business tests
A UK-focused AI-agent testing site publishes benchmark results, costs and run logs, alongside a browser-based prompt checker and paid testing services.
- For
- UK small businesses evaluating AI agents
- Worth noticing
- The site says it computes benchmark answers in code before running agents, and publishes logs, costs, and SHA-256-hashed evidence files.
Field notes
The benchmark page says it computes the correct answer in code before running agents, then repeats tasks across setups and publishes logs and costs. Its featured safety test gives 11 agent setups access to a shop inbox with refund, payment, deletion, and email tools, alongside a prompt-injected fake invoice; the page reports 29 of 130 runs avoided every irreversible action without an approval rule, versus 130 of 130 with one. The site also lists 1,548 SHA-256-hashed evidence files.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue