LinkedIn post — Agent Readiness Report: Cursor

Single post, ~340 words. Polished; opens with the score, closes with a polite tag of Cursor’s leadership.

Agent Readiness Report: Cursor scored 70/100. Level 3 (Agents First).

Second product in the Agent Readiness Reports series to crack Level 3. Cursor is the agent-native IDE — the editor that turned MCP from a protocol document into a default install for working developers — and the docs surface that powers their product reflects the team’s worldview. They sell to humans who write code with agents; their docs read like a contract written for the agents that will read them.

Across three surfaces:

What’s working on docs.cursor.com: a real /AGENTS.md at the docs root (the load-bearing contract artifact promoted to 15 points in v0.2.0 of the rubric), /llms.txt published, an MCP Server Card at /.well-known/mcp-server-card.json, and OAuth-with-PKCE discovery present. Agent-capabilities scores a clean 30/30 — the first surface in the report series to max out that dimension. The homepage hero references MCP, CLI, SDK, and API alongside human onboarding: 10/10 on visibility, the dimension almost everyone fails. Cursor passes because the team building the product lives the way it builds.

What’s missing: no Content-Signal directive on robots.txt anywhere — that’s the cheapest 10 points in the rubric, on the dimension where Cursor scores zero. Markdown content negotiation does not pass. And cursor.com (the marketing root) trips the Agents Without Rules anti-pattern: agent capabilities advertised on the product, no contract file at the root telling an agent how to use them.

Top three fixes:

  1. Lift the docs playbook to cursor.com — /AGENTS.md + MCP Server Card + Content-Signal at the marketing root.
  2. Add Content-Signal to robots.txt on every surface — pick a direction, either earns full credit.
  3. Serve text/markdown when an agent asks for it; docs are already authored in markdown.

The lesson for everyone else: a product team that lives the agent worldview ships the surfaces to match — but only on the surface its team owns directly. Variance from 70 → 30 across subdomains of the same company is this series’ recurring finding. A Level 3 product is Level 3 on every surface an agent might reach.

Full report, raw probe data, and rubric (v0.2.0): https://agentsfirst.dev/reports/cursor/

Third in a bi-weekly Agent Readiness Reports series. Polite tags to Eric Zakariasson on the Cursor docs/DevX team and Michael Truell, Aman Sanger, and Sualeh Asif at Anysphere. Replies and “we just shipped the fix” notes welcome.

#AgentsFirst #MCP #AIAgents #Cursor #DeveloperExperience