Changelog
Since September 2026, cabfy has helped teams assess whether their sites can be understood and operated by AI agents. Here are the major product updates.
September 2026
Tour: see what an agent can actually do
Tour observes a site against a visitor-selected mission, records the steps it followed and shows exactly where the journey stops. A public demo is available; live journeys are included in Growth and Agency, with 10 per month.
Public MCP server
Connect cabfy to Codex, Claude Code, Cursor or any compatible MCP client to launch audits and retrieve results from your working environment.
Agency workspace for teams and clients
The Agency plan brings multiple sites, team members and client access into one shared workspace, with pooled quotas, branded reports and revocable sharing links.
Product repositioning
Cabfy is now an agent-readiness platform: it assesses any website's discoverability, accessibility and operability for AI agents.
General Agent Readiness scan
A platform-agnostic, multi-signal assessment ranks each domain across four maturity levels and identifies the highest-impact levers for improvement.
Public agentic web observatory
Leaderboard, domain score and Data pages make agent-readiness trends publicly accessible, with documented methodology.
Two new measurement pillars
Agent Performance measures an agent's browsing experience; Web Discoverability assesses whether an organisation's official resources can be found.
API, CLI and continuous integration
The general scan is available through the public API and cabfy CLI, with a GitHub App integration to monitor agent-readiness across pull requests.
Multi-site monitoring in the dashboard
The dashboard brings together projects, audit history, score trends and operational recommendations, with expanded visibility for Growth accounts.
New product experience
The landing page, playground and onboarding journey were redesigned around a business goal: quickly understand where to invest to make a site usable by AI agents.
New agent access points
Sites can now document their capabilities with agents.md, progressively expose actions through WebMCP, and declare access conditions for automated crawlers and agents.