API & MCP

Quickstart

A single GET request. No authentication. By default you search FreeSerp Main — main pages only, 20M+ site profiles with LLM summaries (exactly the API you may already use):

curl 'https://freeserp.ai/api.php?q=ai+chatbot&ai_startups=1&sort=went_live&order=desc&size=3'

Add index=web to search the open web with FreeSerp Global — our own 3.1B+ page index. Results are ranked pages: url, title, snippet, language, publication date, and a site card (Domain Rating, category, live date) when we know the domain:

curl 'https://freeserp.ai/api.php?index=web&q=how+to+fine-tune+llama+3&size=3'

Python

import requests
# FreeSerp Main (default): newest AI startups with LLM summaries
r = requests.get('https://freeserp.ai/api.php', params={
    'q': 'ai chatbot', 'ai_startups': 1, 'sort': 'went_live', 'order': 'desc', 'size': 5,
})
for hit in r.json()['results']:
    print(hit['domain'], '—', hit['ai_summary'][:120])

# FreeSerp Global: any page on the web
r = requests.get('https://freeserp.ai/api.php', params={'index': 'web', 'q': 'how to fine-tune llama 3', 'lang': 'en', 'size': 5})
for hit in r.json()['results']:
    print(hit['url'], '—', hit['snippet'][:120])

JavaScript (fetch)

const r = await fetch('https://freeserp.ai/api.php?' + new URLSearchParams({
  index: 'web', q: 'best crm for small business', size: 5,
}));
const { results } = await r.json();
results.forEach(h => console.log(h.url, h.title, h.site?.dr));

Two indexes

Endpoint

Also reachable as /api. Read-only. Responses are cached ~30s at the edge.

Parameters

Response

FreeSerp Global (index=web)

{
  "ok": true,
  "index": "web",
  "query": "how to fine-tune llama 3",
  "total": 97,                      // de-duplicated candidate pool (≤ 100), not an exhaustive count
  "count": 3, "from": 0, "size": 3,
  "sort": "relevance", "ranking": "bm25_trust",
  "filters": { "lang": "en" },
  "took_ms": 412, "engine_ms": 388, "cached": false,
  "results": [
    {
      "url": "https://example.com/blog/fine-tune-llama-3",
      "domain": "example.com",
      "title": "How to fine-tune Llama 3 on your own data",
      "snippet": "… step-by-step guide to fine-tune Llama 3 with LoRA on a single GPU …",
      "language": "en",
      "published_at": "2026-05-14", "crawled_at": "2026-07-12",
      "score": 341.2,
      "site": { "dr": 41, "category": "ai", "ai_categories": ["Code & Dev Tools"], "ai_source": "nextjs", "went_live": "2026-08-22", "summary": "Example is …" },
      "source": { "index": "FreeSerp Global", "captured_at": "2026-07-18T03:05:37Z", "digest": "sha1:…" }
    }
  ],
  "related_apis": { … }
}

site is null when the domain is not in FreeSerp Main. source.captured_at / source.digest identify the exact capture the page came from — reproducible evidence for grounding.

FreeSerp Main (index=sites)

{
  "ok": true,
  "index": "sites",
  "query": "ai chatbot",
  "total": 48213,
  "count": 3,
  "from": 0, "size": 3,
  "sort": "relevance", "order": "desc",
  "filters": { "real_site": "1" },
  "took_ms": 34, "engine_ms": 21,
  "results": [
    {
      "domain": "example.ai",
      "url": "https://example.ai",
      "title": "Example — AI chatbot for support",
      "ai_summary": "Example is an AI chatbot platform that …",
      "category": "ai",
      "ai_categories": ["AI Chatbot & Assistant"],
      "ai_source": "nextjs",
      "dr": 41, "went_live": "2026-08-22", "first_seen": "2026-07-10",
      "tld": "ai", "http_status": 200,
      "content_length": 8123, "html_size": 40213,
      "_score": 3182.4
    }
  ]
}

Sorting — every field

FreeSerp Global: sort=relevance (default) or sort=fresh. FreeSerp Main: pass any of these as sort=, with order=asc|desc:

All fields

FreeSerp Global results always carry: url domain title snippet language published_at crawled_at score site{…} source{index, captured_at, digest}.

Sites index — every field returned per result, and whether you can filter or sort on it:

Field values

The keyword fields you filter on take a fixed set of values. The most useful:

category

Broad heuristic bucket. ai betting/casino (null)  — ai is deliberately wide; prefer ai_startups=1 / ai_categories for real AI products.

ai_source — builder / tech stack

AI-builder signals: ai_likely lovable v0 bolt base44  · classic stacks: wordpress shopify wix webflow nextjs react not_ai

ai_categories — AI sub-niche (multi-valued)

Counted as a real AI startup by ai_startups=1:
AI Agents & Autonomous Code & Dev Tools AI Infrastructure & API AI Automation & Workflows LLM & Prompt Tools AI Search & Answers AI Website Builder No-code / App Builder Image Generation Video Generation Voice & Text-to-Speech Data & Analytics Research & Science Design & UI Chatbot & Assistant

Also present but treated as noise for startup discovery: E-commerce, Directory / Aggregator, Marketing & Ads, Sales & CRM, Lead Gen & Outreach, Finance & Trading, Healthcare & Medical, Education & Tutoring, Other AI, …

Caveats — read this

Data & coverage

FreeSerp Global: 3,132,469,027 pages, built and ranked by us — own crawl, own extraction, own link graph for authority and domain trust. Languages: en 1,256M, ru 228M, de 187M, fr 162M, ja 160M, es 158M, zh 140M, it, pt, pl, nl, tr, cs, vi and more. Every result is reproducible down to the capture (source.captured_at, source.digest).

FreeSerp Main (main pages only): ~20M live sites from our own continuously-refreshed discovery pipeline (newly-registered domains are probed daily; live ones are crawled, LLM-summarised, categorised and DR-scored). Global — every TLD and language. Each record is the homepage: title, extracted text, ai_summary, tech stack, DR and live date. It is a discovery index of what is newly live on the web — not a crawl of every page of every site (that is what FreeSerp Global is for).

Limits, paging & errors

Rate limits

No hard limit and no key — the service is free and open. Please be reasonable (a few requests/second), cache when you can, and identify yourself so we can reach you before a change or ask you to ease off instead of blocking. Responses are edge-cached ~30s.

Paging

Web: from + size ≤ 100 — the engine returns the top-100 candidates for a query; refine the query or add lang/published_from instead of paging deeper. Sites: size ≤ 100 per page; from + size ≤ 10,000. To pull a large set, narrow with filters (date range, ai_categories, dr_min) rather than paging to the cap.

Errors

HTTP 200 with "ok": true on success. On failure you get a non-2xx with an error body:

{ "ok": false, "error": "upstream", "detail": "search_phase_execution_exception" }

Bad params never error — unknown filters are ignored and an invalid sort falls back to relevance, so a query always returns something. The echoed filters object shows exactly what was applied.

Statistics

A live snapshot of the whole index, refreshed every 30 minutes — no heavy query on your side.

{
  "ok": true,
  "generated_at": "2026-09-06T10:16:35+00:00",
  "web_index": { "name": "FreeSerp Global", "pages": 3132469027, "nodes": 18, "languages": { "en": 1256209788, "ru": 228281221, … }, "engine_ready": true },
  "totals":  { "real_sites": 16273795, "all_sites": 21086714 },
  "new":     { "today": 29247, "yesterday": 31800, "last_7d": 232787, "last_30d": 11289030 },
  "ai_startups": { "total": 33429, "today": 72 },
  "by_category": [ {"key":"ai","count":78674}, {"key":"betting/casino","count":806182} ],
  "top_ai_categories": [ {"key":"AI Agents & Autonomous","count":10467}, … ],
  "top_ai_source": [ … ], "top_tld": [ … ],
  "dr": { "avg": 8.4, "min": 0, "max": 99, "histogram": [ … ] },
  "by_day": [ {"key":"2026-09-06","count":29247}, … ],
  "gambling": { "today": {"original":…,"affiliate":…}, "total": {…} },
  "pipeline": { "summary_todo": …, "went_live_today": …, "probe_due": … }
}

Over MCP, the same snapshot is the freeserp_stats tool. The robot also keeps 30 days of half-hourly history in freeserp_stats for trend charts.

Examples & recipes

★ Answer a question from the web (index=web)

curl 'https://freeserp.ai/api.php?index=web&q=how+does+chatgpt+search+choose+sources&size=5'

English results, fresher pages first

curl 'https://freeserp.ai/api.php?index=web&q=best+link+indexing+services&lang=en&fresh=1'

Pages published this year only

curl 'https://freeserp.ai/api.php?index=web&q=gpt-5+pricing&published_from=2026-01-01&size=10'

Grounding for an agent: ask, then read the site card

r = requests.get('https://freeserp.ai/api.php', params={'index': 'web', 'q': 'best backlink indexing service', 'size': 8}).json()
for h in r['results']:
    site = h['site'] or {}
    print(h['url'], '| DR', site.get('dr'), '|', site.get('summary', '')[:80])

★ Discover new AI startups in a niche (default index — recommended combo)

curl 'https://freeserp.ai/api.php?ai_startups=1&ai_categories=AI%20Agents%20%26%20Autonomous&from_date=2026-09-01&sort=went_live&order=desc'

Newest AI sites, most authoritative first

curl 'https://freeserp.ai/api.php?ai=1&sort=dr&order=desc&size=10'

All sites built on Shopify, most authoritative first

curl 'https://freeserp.ai/api.php?ai_source=shopify&sort=dr&order=desc&size=25'

High-DR news sites that went live in the last two weeks

curl 'https://freeserp.ai/api.php?q=news&dr_min=30&from_date=2026-08-16&sort=went_live&order=desc'

Browse by niche, no text query

curl 'https://freeserp.ai/api.php?ai_categories=Image%20Generation&sort=dr&order=desc'

Return full page text, not just the summary

curl 'https://freeserp.ai/api.php?q=ai+chatbot&content=1&content_max=4000&size=3'

Look up one exact domain (record, DR, stack, summary)

Pass the full domain (any TLD) as q — it is boosted to the top; take the first result and confirm results[0].domain matches. Add all=1 so parked/empty domains still resolve.

curl 'https://freeserp.ai/api.php?q=mynewstartup.ai&all=1&size=1'

Is this a real site or parked/dead?

Check real_site and http_status on the returned record (1 = real content, 0 = parked/empty/error).

curl 'https://freeserp.ai/api.php?q=somedomain.io&all=1&size=1'

Find competitors / similar sites in a space

curl 'https://freeserp.ai/api.php?q=vector+database&ai_startups=1&dr_min=10&sort=dr&order=desc&size=25'

Identify yourself (optional)

There is no key and no signup — which also means that when something is about to change, we have no way to reach you. These four fields fix that. Every one is optional, nothing is validated, and none of them affect your results.

curl 'https://freeserp.ai/api.php?q=ai&agent=MyAgent/1.0&[email protected]&project=Acme&website=https://acme.com'

Over MCP, pass the same fields as arguments to any freeserp_search call. Sending them once is enough. What it buys you: we can warn before a breaking change, ask what you are missing, and — if your traffic starts hurting the service — ask you to slow down instead of blocking you. Stored privately next to request counts; never published, never shared.

MCP connector

FreeSerp ships a hosted Streamable-HTTP MCP server — one URL, no install. Tools and their schemas are generated live from the field registry, so they never drift from the API.

A plain GET on that URL returns the live tool catalog (server, transport, tool_count, and every tool with its description) — same shape your client discovers over JSON-RPC:

curl 'https://freeserp.ai/mcp'

Add it to Claude (Desktop / Code / any MCP client)

Point the client at the remote endpoint via mcp-remote:

{
  "mcpServers": {
    "freeserp": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://freeserp.ai/mcp"]
    }
  }
}

Restart the client — your agent now has these tools:

Raw JSON-RPC

Talk to it directly — initialize, then tools/list / tools/call:

# discover tools
curl -s 'https://freeserp.ai/mcp' -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

# run a search
curl -s 'https://freeserp.ai/mcp' -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{
        "name":"freeserp_search",
        "arguments":{"q":"ai chatbot","ai_startups":true,"sort":"went_live","size":5}}}'

# FreeSerp Global: any page on the web
curl -s 'https://freeserp.ai/mcp' -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{
        "name":"freeserp_search",
        "arguments":{"index":"web","q":"how to fine-tune llama 3","lang":"en","size":5}}}'