Vercel Launches Is Agentic to Measure AI Agent Readiness
Vercel has introduced Is Agentic, a free tool and API suite designed to audit how effectively AI agents can discover and use website content.
Vercel's Is Agentic platform provides developers with a structured methodology to audit how AI models interact with their websites.
- Vercel launched Is Agentic, a tool providing a 0–100 score for a website's compatibility with AI agents.
- The platform includes a CLI, a public REST API, and support for the Model Context Protocol (MCP).
- Audits are based on research by Ora and include over 100 checks covering document structure, server-side rendering, and API discovery.
Evaluating the Agentic Web
Vercel Developers have announced the launch of Is Agentic, a technical auditing tool designed to quantify how easily AI agents can navigate, interpret, and interact with public websites. As autonomous agents become more prevalent in software development and browsing workflows, this tool aims to provide developers with a standardized benchmark for "agent readiness." The tool is backed by research from Ora and provides a numeric score from 0 to 100 based on more than 100 individual checks.
The service evaluates the public-facing aspects of a URL to determine if an agent can successfully retrieve and understand its content. According to the documentation, the scoring methodology prioritizes "Essential" checks, which include baseline requirements such as server-rendered content, standard HTTP behavior, clear document hierarchy, and usable UI controls. Websites are not penalized for lacking specific features like GraphQL endpoints or commerce surfaces if those features are not relevant to the site's purpose; instead, these fall under "Recommended" checks that only activate when evidence of such interfaces is detected.
Technical Architecture and Scoring
The Is Agentic platform operates through several interfaces intended for both human developers and the agents themselves. On the web interface, users can enter a URL to receive a detailed report including a numeric score, a visualization of an agent’s journey through the site, and specific technical recommendations. For example, recent scores listed on the platform show Vercel’s own domain at 89/100, while others like Anthropic (67/100) and Shopify (66/100) demonstrate the current variability in how modern sites handle non-human traffic.
Every report includes evidence observed during a scan. If a site fails a check, the tool provides a concrete recommendation that can be integrated into a development workflow. The platform also tracks "Observed Tasks," showing the friction points an agent encountered during a test run. While these runs are illustrative rather than a direct component of the numeric score, they serve as supporting evidence for why a site may be difficult for an AI to parse.
The CLI and API Ecosystem
For automated workflows, Vercel has released an official CLI via the is-agentic package on npm. By running npx is-agentic [domain], developers can retrieve the latest completed report directly in their terminal. The CLI supports a --json flag, allowing for structured data output that can be consumed by CI/CD pipelines or other scripts.
The public REST API, currently at version v1, follows standard REST conventions and provides machine-readable results. The API includes robust error handling using RFC 9457 problem details, providing stable error codes like rate_limit_exceeded or report_not_found. Rate limits are set at 120 requests per client IP per minute, with responses utilizing IETF RateLimit headers to inform clients of their remaining quota. Furthermore, the API supports content negotiation, allowing clients to request a compact Markdown representation of a report by setting the Accept: text/markdown header.
Model Context Protocol Integration
In a significant move for agent-to-agent interoperability, Is Agentic supports the Model Context Protocol (MCP). This allows AI models to use the Is Agentic suite as a toolset within their own execution environments. The MCP server exposes three read-only tools: is_agentic_get_report for scan results, is_agentic_get_methodology for interpreting scores, and is_agentic_get_developer_docs for technical guidance.
This integration enables an agent to evaluate the "agentic" nature of a target website before attempting to interact with it, effectively allowing agents to self-diagnose potential hurdles in their browsing tasks. The MCP implementation includes support for "MCP Apps," which can render an interactive score card directly within a model's host interface if supported.
Availability and Future Outlook
Is Agentic is currently free to use, with no subscription fees or per-report charges for the public website, CLI, or API. Vercel emphasizes that while the tool provides a prioritized technical review, it does not guarantee that every AI model will succeed on a high-scoring site, as factors like private pages, authentication, and bot defenses remain outside the scope of a public scan. By providing a stable, machine-readable URL for every report, Vercel aims to create a long-term reference point for developers to compare their site’s accessibility to AI agents over time.
Enjoyed this?
Get more posts like this delivered to your inbox.
