generated: '2026-09-02'
method: derived
source: mcp/pagesnap-mcp-tools.json + a2a/pagesnap-agent-card.json + openapi/pagesnap-openapi.json
note: >-
  Both the MCP tool list and the A2A skill list were fetched live and unauthenticated, and the
  OpenAPI 3.1 document is the provider's own generated contract, so every binding below is a real
  name-to-operationId match rather than a guess. Confidence is high wherever the tool's input
  properties are a subset of the REST operation's parameters.
surfaces:
  openapi:
    path: openapi/pagesnap-openapi.json
    url: https://pagesnap.142-93-197-141.sslip.io/openapi.json
    operations: 66
    gated: false
  mcp:
    url: https://pagesnap.142-93-197-141.sslip.io/mcp
    tools: 11
    gated: false
    note: Anonymous tools/list returned full inputSchema for all 11 tools.
  a2a:
    url: https://pagesnap.142-93-197-141.sslip.io/a2a
    skills: 6
    gated: false
  graphql: null
crosswalk:
  - tool: read_url
    a2a_skill: read_url
    category: content
    rest: [readUrl, readUrlPost, readShortcut, readUrlX402]
    binding: direct
    confidence: high
    note: Tool inputs url/format/render/lang/max_chars map onto /v1/read query parameters.
  - tool: screenshot_url
    a2a_skill: screenshot_url
    category: content
    rest: [screenshotUrl, screenshotUrlPost, screenshotShortcut, screenshotUrlX402]
    binding: direct
    confidence: high
  - tool: pdf_url
    a2a_skill: pdf_url
    category: content
    rest: [pdfUrl, pdfUrlPost, pdfUrlX402]
    binding: direct
    confidence: high
  - tool: get_meta
    a2a_skill: get_meta
    category: content
    rest: [getMeta, getMetaX402]
    binding: direct
    confidence: high
  - tool: extract_structured
    a2a_skill: extract_structured
    category: content
    rest: [extractStructured, extractStructuredX402]
    binding: direct
    confidence: high
  - tool: crawl_site
    a2a_skill: crawl_site
    category: crawl
    rest: [crawlSite, crawlSiteX402, getCrawlJob, getCrawlJobResult]
    binding: composite
    confidence: high
    note: >-
      The tool hides the 202 job handshake. Effective limits over 25 return a job_id at REST;
      the MCP tool polls and returns the finished Markdown.
  - tool: generate_llms_txt
    category: crawl
    rest: [generateLlmsTxt, getCrawlJob, getCrawlJobResult]
    binding: composite
    confidence: high
  - tool: create_monitor
    category: monitors
    rest: [createMonitor]
    binding: direct
    confidence: high
    note: Key required. Returns webhook_secret once, as the REST 201 does.
  - tool: list_monitors
    category: monitors
    rest: [listMonitors]
    binding: direct
    confidence: high
  - tool: usage
    category: account
    rest: [getAccount, getUsage]
    binding: direct
    confidence: high
  - tool: upgrade_plan
    category: billing
    rest: [createCheckout]
    binding: direct
    confidence: medium
    note: >-
      Maps to POST /v1/checkout. The tool returns a pay URL; polling getCheckout is left to the
      caller, so this is not a full composite of the checkout flow.
mcp_only: []
a2a_only: []
rest_only:
  - operations: [batchRead]
    reason: Batch read has no MCP tool; agents loop read_url instead.
  - operations: [renderHtml, renderHtmlX402]
    reason: Supplying raw HTML is a REST/x402-only surface.
  - operations: [diffUrls]
    reason: >-
      Two-URL diff is REST-only as a tool, though the MCP prompt compare_pages composes read_url
      twice to cover the same job.
  - operations: [getMonitor, deleteMonitor, checkMonitor, pauseMonitor, resumeMonitor]
    reason: Monitor lifecycle beyond create/list is REST-only; MCP exposes no destructive monitor tool.
  - operations: [createKey, verifyLinkGrant]
    reason: Key issuance and the link-grant program are account operations with no tool.
  - operations: [getCheckout, getReceiptHtml, getReceiptPdf, getPayPageStatus]
    reason: Payment polling and receipts are REST-only.
  - operations: [createSupportTicket, getSupportTicket, addSupportTicketMessage]
    reason: Support ticketing is REST-only.
  - operations: [getPublicStats, getPublicStatus, healthcheck]
    reason: Status and stats are REST-only; MCP callers read pagesnap://usage instead.
  - operations: [agentCard, agentJwks, apiCatalog, securityTxt, x402Manifest, ardManifest, aiCatalog, llmsTxt, llmsFullTxt, openapiJson, openapiYaml, humansTxt, poweredByBadge, ogImage, getTemporaryResult]
    reason: Discovery, documentation and static assets; not tool-shaped.
  - operations: [mcpGet, mcpPost, mcpDelete, a2a]
    reason: The protocol transports themselves, described inside their own OpenAPI.
coverage:
  mcp_tools: 11
  a2a_skills: 6
  rest_operations: 66
  tools_bound_to_rest: 11
  tools_without_rest: 0
  rest_without_tool: 55
  note: >-
    MCP is a deliberate subset, not a projection gap: it covers the six content skills plus crawl
    automation, monitors and account/billing, and omits the discovery, support, receipt and
    status surfaces that an agent reads once rather than calls repeatedly.