Free · no signup · MCP + REST
Check every link before your AI cites it.
LinkVet is a tiny API for AI agents. Send a URL, get back JSON: is it live, a 404, or blocked? Where does it really end up after redirects? What's the title, price, and publish date? And one field that matters most: safe_to_cite.
// result appears here
Stop hallucinated links
Models invent plausible URLs. One call tells you if the page really exists before a person clicks it.Cite the right URL
Follows up to 5 redirects and returnsfinal_url, so you link to where the page lives now.Facts in one shot
Title, description, price (JSON-LD/Open Graph), published and modified dates, canonical URL.Batch + cached
Up to 25 URLs per call. Results are cached ~12 hours, and cached hits are free and instant.REST API
Base URL: https://linkvet.pages.dev. No API key. CORS enabled. Spec: /openapi.json.
curl "https://linkvet.pages.dev/api/check?url=https://example.com"
curl -X POST https://linkvet.pages.dev/api/check \
-H "content-type: application/json" \
-d '{"urls":["https://example.com","https://example.com/missing-page"]}'
Single URL returns one object. A urls list (max 25) returns {"count":N,"results":[...]} in input order. Sample result:
{
"input": "https://example.com",
"url": "https://example.com/",
"status": "live",
"http_status": 200,
"final_url": "https://example.com/",
"redirected": false,
"redirect_chain": [],
"title": "Example Domain",
"description": null,
"site_name": null,
"canonical": null,
"price": null,
"published": null,
"modified": null,
"last_modified_header": null,
"content_type": "text/html",
"safe_to_cite": true,
"reason": "live_200",
"elapsed_ms": 143,
"checked_at": "2026-11-22T03:40:00.000Z",
"cache_expires_at": "2026-11-22T15:40:00.000Z",
"cached": false
}
| status | meaning | safe_to_cite |
|---|---|---|
live | 2xx response. Still marked unsafe if the title looks like a soft-404 or a parked domain. | usually true |
404 | 404 or 410: the page is gone | false |
blocked | 401/403/429 or a bot challenge. The page may exist, but it couldn't be verified. | false |
timeout | no response within 8 seconds | false |
error | DNS failure, 5xx, too many redirects | false |
invalid / blocked_ssrf | not an http(s) URL, or it points at a private/internal address (rejected, never fetched) | false |
MCP server
Remote MCP endpoint (streamable HTTP, no auth): https://linkvet.pages.dev/mcp. Tools: check_url and check_urls.
{
"mcpServers": {
"linkvet": { "type": "http", "url": "https://linkvet.pages.dev/mcp" }
}
}
Claude Code: claude mcp add --transport http linkvet https://linkvet.pages.dev/mcp
Clients that only speak stdio can use npx mcp-remote https://linkvet.pages.dev/mcp.
Pricing
| Tier | Price | Details |
|---|---|---|
| Free | $0 | 100 fresh checks per day per IP. Cached results don't count. |
| Pay per call | about $0.001 per check | No account needed. Agents pay inline with x402 (USDC on Base, HTTP 402) or L402 (Bitcoin Lightning). Coming soon: until it's switched on, requests over the free limit get HTTP 429. |
Why agents use it
Put this line in your system prompt: "Before citing any URL, call LinkVet check_url (or check_urls for several). Only cite links with safe_to_cite=true, and cite final_url."
Safety
Only http and https on standard ports are fetched. Localhost, private, link-local, and reserved IP ranges (IPv4 and IPv6) are rejected, including hostnames that resolve to them and redirects that point to them. Credentials in URLs are refused. Only the first 256 KB of HTML is read.