generated: '2026-09-19'
method: derived
source: >-
  Derived by aligning the LIVE MCP tools/list from https://agent-ready.dev/api/v1/mcp and
  https://agent-ready.dev/api/apps/mcp (both answered anonymously; saved to mcp/agent-ready-dev-mcp-tools.json and
  mcp/agent-ready-dev-mcp-apps-tools.json) and the three A2A skills in a2a/agent-ready-dev-agent-card.json with
  the provider's OpenAPI 3.1.0 at openapi/agent-ready-dev-openapi.yml (8 operations). The provider's own
  /.well-known/agents.json (Wildcard v0.1.0) binds the same flows to the same operationIds, which corroborates
  every high-confidence row below.
purpose: >-
  Make the MCP tool / A2A skill the first-class discovery unit and bind it to its backing REST operation. Agent
  Ready's three surfaces are projections of one scanner: MCP and A2A expose the same three names, REST exposes
  those plus a list endpoint, a public MCP-server grader and a paid x402 lane that neither agent surface carries.
surfaces:
  rest_openapi: openapi/agent-ready-dev-openapi.yml            # 8 operations, 19 schemas; source in openapi/_original/
  mcp: https://agent-ready.dev/api/v1/mcp                        # tools/list PUBLIC (observed); tools/call for scan_site/get_scan needs Bearer key
  mcp_apps: https://agent-ready.dev/api/apps/mcp                 # tools/list and tools/call PUBLIC, no auth
  a2a: https://agent-ready.dev/api/v1/a2a                        # skills from the served agent card; JSON-RPC endpoint not called by this pipeline
  graphql: null
crosswalk:
- tool: scan_site
  a2a_skill: scan_site
  category: scan
  rest: [startScan, getScan]
  binding: rest
  confidence: high
  note: >-
    The tool's inputSchema (url required, pageLimit 1-2000 optional) is exactly StartScanRequest. The tool
    description says a long scan returns an id to poll — i.e. the server performs POST /api/v1/scans (202 +
    Location) and, when time allows, follows with GET /api/v1/scans/{id}; the outputSchema is the Scan shape
    (vercelScore, llmstxtScore, accessibilityScore, protocolResults …). On the Apps endpoint pageLimit is not
    exposed. agents.json binds the same pair as its start_scan_flow.
- tool: get_scan
  a2a_skill: get_scan
  category: scan
  rest: [getScan]
  binding: rest
  confidence: high
  note: 'inputSchema {id} is the getScan path parameter; the A2A card maps the skill to the tasks/get method.'
- tool: ask
  a2a_skill: ask
  category: search
  rest: [askPost, askGet]
  binding: rest
  confidence: high
  note: >-
    inputSchema {q, itemType, mode} mirrors AskRequest (query, itemType, mode) on POST /api/v1/ask; GET
    /api/v1/ask carries the same fields as query parameters plus stream. Public on every surface.
mcp_only: []
rest_only:
- operation: listScans
  capability: scan history
  reason: Cursor-paginated list of the key's past scans (Pro); no MCP tool or A2A skill lists history — the CLI exposes it as `agent-ready list`.
- operation: scanMcp
  capability: MCP server grading
  reason: >-
    Public POST /api/v1/scan/mcp grades a live remote MCP server (mcpScore 0-100). Not an MCP tool on either
    hosted endpoint; the CLI exposes it as `agent-ready mcp-scan <endpoint>`.
- operation: x402ScanChallenge
  capability: agent payment
  reason: GET /api/x402/scan returns the x402 v2 / MPP HTTP 402 challenge. Payment is a REST-only lane; the provider's docs say the MCP server, CLI and SDK never use it.
- operation: x402Scan
  capability: agent payment
  reason: POST /api/x402/scan runs a scan paid with an X-PAYMENT header (USDC on Base). REST-only, no account.
coverage:
  mcp_tools_named: 3
  mcp_tools_bound: 3
  mcp_only: 0
  a2a_skills_named: 3
  a2a_skills_bound: 3
  rest_operations_total: 8
  rest_operations_with_tool: 4
  rest_only: 4