generated: '2026-09-19'
method: probed
status: published
source: https://api.xguardgate.com/mcp
docs:
- https://xguardgate.com/developers
- https://github.com/moelayyan90/XGuard/blob/main/llms-install.md
- https://xguardgate.com/.well-known/mcp/server-card.json
summary: >-
  XGuard operates ONE remote MCP server at https://api.xguardgate.com/mcp, on the same host as its OpenAPI,
  its A2A JSON-RPC endpoint and its x402 facilitator relay. It is Streamable HTTP, speaks MCP protocol version
  2026-07-28, identifies itself as serverInfo {xguard-universal-paid-secretless-gateway, 5.1.0}, and answers
  initialize and tools/list anonymously with exactly three tools, each carrying a real JSON Schema inputSchema
  and MCP tool annotations (readOnlyHint / destructiveHint / idempotentHint / openWorldHint). The initialize
  instructions and the provider's README both say the live tools/list is deliberately small: the paid-outcome
  chokepoint is xguard_execute, and the older explicit tool names that llms-install.md still lists
  (xguard.web.fetch, xguard_secretless_egress, xguard_proofrail, ...) are "compatibility references, not the
  default catalog". A paid tool call returns an HTTP 200 JSON-RPC result with isError: true and the x402
  PaymentRequired challenge in both structuredContent and content[0].text; a funded x402 MCP client retries the
  identical arguments with params._meta["x402/payment"] and receives settlement in
  result._meta["x402/payment-response"]. Authorship is not in question: the tool names, the serverInfo name,
  the canonical-identity block in every response header set (x-xguard-canonical-mcp) and the MCP Registry
  entry io.github.moelayyan90/xguard-control-plane all name this provider.
server:
  name: xguard-universal-paid-secretless-gateway
  title: XGuard Universal Paid AI Agent + Secretless Gateway
  version: 5.1.0
  transport: streamable-http
  url: https://api.xguardgate.com/mcp
  protocol_version: '2026-07-28'
  capabilities:
    tools: {listChanged: false}
    resources: {subscribe: false, listChanged: false}
    prompts: {listChanged: false}
  registry:
    name: io.github.moelayyan90/xguard-control-plane
    manifest: https://github.com/moelayyan90/XGuard/blob/main/server.json
    schema: https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json
    note: server.json in the repository says version 5.1.1 while the live server, the OpenAPI, the agent card and every x-xguard-version response header say 5.1.0 on 2026-09-19.
  server_card: mcp/xguardgate-com-mcp-server-card.json
  server_card_source: https://api.xguardgate.com/.well-known/mcp/server-card.json
  tools_list: mcp/xguardgate-com-mcp-tools.json
  listings:
  - Official MCP Registry (server.json, io.github.moelayyan90/xguard-control-plane)
  - glama.json maintainer manifest in the repository
  - agent-plugins.org mcp.json / plugin.json manifests in the repository
deployment:
  mode: remote
  endpoint: https://api.xguardgate.com/mcp
  auth: none
  verified: probed
  install_configs:
  - client: Claude Code
    url: https://xguardgate.com/install/claude-code.json
    config: '{"mcpServers":{"xguard":{"type":"http","url":"https://api.xguardgate.com/mcp"}}}'
  - client: Cursor
    url: https://xguardgate.com/install/cursor.json
    config: '{"mcpServers":{"xguard":{"url":"https://api.xguardgate.com/mcp"}}}'
  - client: VS Code
    url: https://xguardgate.com/install/vscode.json
    config: '{"servers":{"xguard":{"type":"http","url":"https://api.xguardgate.com/mcp"}}}'
  - client: Codex
    url: https://xguardgate.com/install/codex.toml
    config: '[mcp_servers.xguard]\nurl = "https://api.xguardgate.com/mcp"'
  note: >-
    A hosted HTTPS endpoint an MCP client POSTs to directly; llms-install.md says "Do not clone, build, or run a
    local process merely to connect it" and no stdio package or npx/uvx install is documented. auth is "none" in
    the connection sense: initialize and tools/list need no credential, the RFC 9728 document at
    /.well-known/oauth-protected-resource/mcp declares bearer_methods_supported [] and oauth_supported false,
    and /.well-known/oauth-authorization-server returns a deliberate 404 ("XGuard does not advertise a
    fictitious OAuth issuer"). Paid tool calls are gated by x402 PAYMENT (USDC on Base, eip155:8453), not by an
    account; installing the server supplies no wallet.
probes:
- {method: POST, url: 'https://api.xguardgate.com/mcp', rpc: initialize, http_status: 200, fetched: '2026-09-19', result: 'protocolVersion 2026-07-28; serverInfo xguard-universal-paid-secretless-gateway 5.1.0; capabilities tools/resources/prompts'}
- {method: POST, url: 'https://api.xguardgate.com/mcp', rpc: tools/list, http_status: 200, fetched: '2026-09-19', result: '3 tools with inputSchema and annotations (saved verbatim to mcp/xguardgate-com-mcp-tools.json)'}
- {method: GET, url: 'https://api.xguardgate.com/.well-known/mcp/server-card.json', http_status: 200, fetched: '2026-09-19', result: 'server card naming endpoint, transport streamable-http, authentication {required:false, schemes:[]}, the same 3 tools and 4 capabilities with input/output schemas (saved verbatim)'}
- {method: GET, url: 'https://api.xguardgate.com/.well-known/oauth-protected-resource/mcp', http_status: 200, fetched: '2026-09-19', result: 'resource https://api.xguardgate.com/mcp, bearer_methods_supported [], oauth_supported false'}
tool_count: 3
tools:
- name: xguard_discover
  category: discovery
  description: Inspect executable outcomes, prices and examples. Free; no setup.
  input: no parameters (additionalProperties false)
  annotations: {readOnlyHint: true, openWorldHint: false}
  rest: GET /v1/capabilities, GET /v1/pricing
- name: xguard_execute
  category: execution
  description: >-
    Extract multi-page evidence, compare structured product offers, or merge feeds. intent:'demo' is free.
    Paid work returns an x402 PaymentRequired tool result; a funded x402 MCP client authorizes and retries with
    params._meta['x402/payment'], preserving the signed quote.
  input: intent (string|object), capability (enum extract-preview|web-extraction|product-offers|feed-digest), url, urls[<=3], sources[<=3], action, desired_action, html (<=12288 chars), limit (1-30), max_age_seconds (0-60), query, testnet
  annotations: {readOnlyHint: false, destructiveHint: false, idempotentHint: false, openWorldHint: true}
  rest: POST /v1/execute (operationId xguardExecute)
- name: xguard_get_result
  category: recovery
  description: Recover a paid outcome using its payment identifier and original signed quote; never charges or executes again. The quote is a bearer recovery credential.
  input: payment_identifier, quote
  annotations: {readOnlyHint: true, openWorldHint: false}
  rest: GET /v1/results/{payment_identifier} with header X-XGuard-Quote
payment:
  protocol: x402 v2
  network: eip155:8453 (Base mainnet); Base Sepolia via testnet:true
  asset: USDC 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913
  challenge: isError tool result carrying PaymentRequired in structuredContent and content[0].text
  retry: identical arguments plus params._meta["x402/payment"]
  settlement: result._meta["x402/payment-response"]
  official_client: '@x402/mcp (https://docs.cdp.coinbase.com/x402/buyer/mcp-payments), as linked from xguardgate.com/developers'
crosswalk: mcp/xguardgate-com-tool-crosswalk.yml