# IPIPAI — IP Intelligence & Network Diagnostics for AI Agents

> Single-retrieval index: this file. Full content: https://ipipai.com/llms-full.txt
> Agent instructions: https://ipipai.com/agents.md
> 中文站: https://ipipai.com/index-zh.html（API 加 `?lang=zh`）

IPIPAI (https://ipipai.com) is a free IP intelligence and network-diagnostic
platform: IP geolocation/ASN/ISP/type/risk for any IPv4/IPv6, proxy/VPN/Tor detection (multi-source: Tor Project official exits, VPN provider IPs,
  OpenProxyDB cross-validation), crawler verification (FCrDNS: verified/impostor/unverifiable),
  rDNS-based ispClass hints
detection, streaming & LLM access guidance, ping/DNS/whois, proxy-node
parsing, BGP change feed. No registration for basic use; machine-readable JSON
API + two MCP servers for AI models.

> **Network perspective**: All web UI tools (ping, streaming unlock, speed test)
> run **from the visitor's browser** — they measure the *user's* network, not
> our server. The server only serves the page. REST APIs below (e.g.
> `POST /api/ping`, `GET /api/dns-resolve`) run server-side and are for
> programmatic use.

## Two MCP endpoints (pick one)

| Endpoint | Tools | Best for |
|---|---|---|
| `POST https://ipipai.com/mcp` | 11 | IP intelligence-first (English tool descriptions) |
| `POST https://mcp.ipipai.com/mcp` | 16 | Network-ops-first (Chinese 意图路由 `when_to_use`, BGP feed, node parsing, globalping) |

Both speak JSON-RPC 2.0 / Streamable HTTP, protocol `2025-06-18`, no key,
60 req/min. Both proxy the same backend: MCP → REST (`/api/*`) → Python
compute engine (compute-on-query over RIR stats, geofeeds, cloud ip-ranges —
no third-party DB in the query path). Results match https://ipipai.com UI.

## Intent → tool routing

- "Where is 1.2.3.4 / is it proxy/VPN/datacenter/mobile/Tor / risk?" →
  `lookup_ip` (main) or `ip_lookup` (sub). Primary IP tool.
- "What's my IP / current egress IP?" → `my_ip` (sub) or `GET /api/myip`.
- "Can I access ChatGPT / Claude / Gemini from here?" → `llm_detection`
  (sub; browser-side test, returns guide) or `GET /api/detect/llm`.
- "Can this IP unlock Netflix / Disney+?" → `check_service_access` (main,
  needs `service`) or `streaming_detection` (sub, browser-side guide).
- "Can this proxy node (vmess/vless/ss/trojan) unlock X?" →
  `streaming_unlock` (sub, needs node URI).
- "Rate this VPS/node (score, grade, stars)?" → `rate_ip_node` (main).
- "Parse this subscription/node link" → `parse_node` (sub).
- "Ping / traceroute / latency?" → `ping_host`, `trace_route` (main) or
  `ping`, `traceroute`, `latency_test` (sub).
- "Test from many countries" → `globalping` (sub).
- "BGP route changes for ASN/IP?" → `asn_changes`, `asn_status` (sub).
- "Whois / DNS / MAC vendor / website censorship check?" →
  `lookup_domain_whois`, `resolve_dns`, `lookup_mac_vendor`,
  `check_website_availability` (main) or `whois_lookup`, `dns_resolve` (sub).
- "Is my IP leaking (WebRTC/DNS)?" → `ip_leak_detection` (sub, guide).

Main-site tool list: https://ipipai.com/mcp-tools.json · Docs: https://ipipai.com/mcp-docs.html

## recommendations block (AI follow-ups, default on)

The 4 IP tools (`lookup_ip`, `check_ip_reputation`, `check_service_access`,
`rate_ip_node`) return a top-level `recommendations` object:
`relatedRanges[]` (recently changed sibling ranges), `asnSummary{torExits,
vpnL2Plus}` (same-ASN reputation), `followUps[]` (up to 3 real follow-up
queries as `{tool, ip, reason}`). Disable per-call with
`"recommendations": false`.

## Core REST endpoints (direct HTTP, JSON)

- `GET /api/myip` → caller IP
- `GET /api/detect/{ip}?lang=en` → full intelligence (see fields below)
- `GET /api/detect/llm|streaming|leak` → browser-test guides (JSON)
- `GET /api/streaming/list` · `POST /api/ping` `{target,count}` (server-side;
  web UI does browser-side ping instead) ·
  `POST /api/globalping/ping`
  `{target,locations,packets}` · `GET /api/whois/{q}` ·
  `GET /api/dns-resolve?domain=` (server-side DNS) · `GET /api/mac-lookup/{mac}` ·
  `GET /api/availability-check` · `POST /api/parse-node` `{input}` ·
  `GET /api/health`
- Web UI browser-side tests (user's network): ping-test.html, unlock-test.html
  (streaming), speed-test.html — all fetch directly from visitor's browser
- AI assistant: `POST /api/ai-chat` `{messages[],model?}` (OpenAI-compatible,
  20 req/min) · `GET /api/ai-models` · `GET /api/ai-health` ·
  `GET /api/ai-hot-ranges`
- Blog: `GET /api/blogs` (paginated) · `GET /api/blogs/:slug`
- ASN: `GET /api/asn/status` · `GET /api/asn/changes`
- Network: `POST /api/latency` `{target,rounds}` ·
  `POST /api/globalping/mtr|trace` `{target,locations}` ·
  `GET /api/myip/public`
- Streaming: `GET /api/streaming/check/:id` · `POST /api/streaming/unlock`
  `{nodeUri,service}`
- Site: `GET /api/site-config` · `GET /api/version`
- Key tier: `GET /api/v1/detect/{ip}` with `X-API-Key` header (120 req/min)

## Key /api/detect response fields

`ip`, `countryCode` (ISO), `country/countryEn`, `region/regionEn`,
`city/cityEn`, `lat/lon`, `timezone`, `isp`, `org`, `as` ("AS15169 Google LLC"),
`asn`, `ipTypeCode` (CDN/CLD/MOB/DCH/RES/EDU/GOV/BIZ),
`ispClass` (mobile/residential/datacenter/enterprise/institutional/unknown),
`isProxy/isHosting/isMobile/isTor/isCDN/isCloud/isBusiness`,
`vpn_confidence` (0=none,1=rdns/org,2=+ports,3=+config),
`carrier_name`, `mcc/mnc`, `category/categoryEn`, `usageType`,
`riskScore` (0-100), `riskLevel`, `riskColor` (hex), `riskFactors[]`,
`uncertaintyRadiusKm` + `uncertaintyTierKm` (5/10/20/50/100/500/5000),
`dataSource` ("computed"), `engineVersion` ("2.0-compute"),
`schemaVersion` (2), `conclusion{verdict,verdictColor,verdictIcon,suggestion}`,
`recommendations{...}` (MCP IP tools / documented above).

## Rate limits & errors

- `/mcp`, `/api/detect`: 60 req/min (free, no key). 429 → wait 60s, retry.
- MCP `tools/call` returns `isError`; unknown tool → `-32601`.
- Detection is probabilistic; data cached ~24h.

## GEO / AEO articles (bilingual, answer-first)

Question-led guides with FAQ schema, one canonical URL per language
(`/blog/{slug}` = Chinese, `/blog/{slug}/en` = English):

- Public IP check: five ways and which to trust —
  https://ipipai.com/blog/public-ip-check-methods ·
  EN https://ipipai.com/blog/public-ip-check-methods/en
- How accurate is IP geolocation (country vs city vs street) —
  https://ipipai.com/blog/how-accurate-is-ip-geolocation ·
  EN https://ipipai.com/blog/how-accurate-is-ip-geolocation/en
- Bandwidth vs throughput: what a speed test really measures —
  https://ipipai.com/blog/bandwidth-vs-throughput-test ·
  EN https://ipipai.com/blog/bandwidth-vs-throughput-test/en
- IP geolocation MCP server for Claude and Cursor —
  https://ipipai.com/blog/ip-geolocation-mcp-server ·
  EN https://ipipai.com/blog/ip-geolocation-mcp-server/en

## Key pages

- Home EN https://ipipai.com/ · ZH https://ipipai.com/index-zh.html
- Blog hub https://ipipai.com/blog.html (20 curated guides; full index: https://ipipai.com/sitemap_agentic_discovery.xml)
- Docs https://ipipai.com/docs.html · Tech doc https://ipipai.com/tech-doc.html
- `ip/{ip}` · `asn/{asn}` · `country/{cc}` detail pages · `unlock/{service}`
- Privacy https://ipipai.com/privacy.html · About https://ipipai.com/about.html