agent-ready-mcp
Agent Ready MCP server
The Agent Ready (agent-ready.dev) MCP server exposes the ~71-check agent-readability scan (plus a separate 23-check accessibility sub-score) as MCP tools. Install once and your agent can run scans, fetch previous results, and search the Agent Ready docs through MCP-native tool calls — no HTTP wiring required.
This skill covers two distinct phases:
- Install and configure the server in the user's MCP client (Claude Desktop, Claude Code, Cursor, Cline, Continue, Goose, or any other streamable-HTTP / stdio MCP host).
- Use the tools and prompts the server exposes once it's running.
The server is published to npm as agent-ready-mcp; source at https://github.com/mlava/agent-ready-mcp. If the user has no MCP client and just wants REST calls (curl / fetch / requests), use the agent-ready-api skill instead.
Step 1: Get an API key (optional for scan_site)
scan_site works with no key at all, on the free anonymous tier (3 scans per 30 days per IP, 25-page depth) — you can skip straight to Step 2 and omit the env block. ask is always public too, no key needed.
get_scan always needs a Pro account key — scan history is account-scoped, so there's no anonymous equivalent. A Pro key also unlocks deeper scan_site runs (250 pages) and higher volume. Issue one at https://agent-ready.dev/dashboard/api-keys (sign up at https://agent-ready.dev/pricing if needed). Keys begin with ar_live_….