Search APIs return SEO'd blue links made for human eyeballs. Rummage returns answers your agents can actually use — cross-encoder reranking, authority boosting, junk filtering, freshness awareness, and every result tagged with why it ranked, so the agent can decide what to trust.
Get a key with one call — no card, no dashboard. 100 searches a month free.
curl -X POST https://ai-search-api.gamebox-cloud-staging.workers.dev/v1/keys \
-H "Content-Type: application/json" \
-d '{"email":"you@example.com"}'
Returns {"key": "aisk_…", "quota_monthly": 500}. Save the key now — it's shown once and never stored in plaintext.
curl -X POST https://ai-search-api.gamebox-cloud-staging.workers.dev/v1/search \
-H "Authorization: Bearer aisk_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"query":"how do vector databases handle hybrid search?","max_results":5}'
The librarian pipeline: upstream search → embeddings → cross-encoder rerank → quality signals on every result.
"quick" (default, 1 quota unit) or "deep" — queries both upstreams and merges, for research-grade questions. Costs 2 quota units.{
"query": "how do vector databases handle hybrid search?",
"upstream": "tavily",
"rerank": "cross-encoder",
"results": [
{
"title": "…",
"url": "…",
"snippet": "…",
"relevance": 0.87,
"signals": [
"source:docs",
"authority:official-docs",
"match:title-strong"
]
}
]
}
Every result carries signals — the why behind its rank, in plain tags any agent can read. relevance orders the result set; treat it as relative (0.92 vs 0.87 is a nudge, not a landslide), and lean on the signals for trust calls.
docs · news · forum · code · video · reference · web — weigh source kinds per task.official-docs (docs subdomains, Wikipedia/GitHub/Stack Overflow) or trusted-tld (.gov/.edu/.org).Clean, readable text from up to 5 URLs — no nav bars, no cookie banners, no ad slots.
curl -X POST https://ai-search-api.gamebox-cloud-staging.workers.dev/v1/extract \
-H "Authorization: Bearer aisk_YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"urls":["https://example.com/article"]}'
Give your agent native search tools over Streamable HTTP (JSON-RPC 2.0). Same key, same quota, same pipeline.
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": {
"name": "search",
"arguments": { "query": "…", "depth": "deep", "compact": true }
}
}
Tools: search and extract. Point any MCP client at /mcp with your Bearer key.
{"plan":"pro"} or {"plan":"scale"} — returns a secure Square checkout link.Upgrading applies the new quota the moment payment completes, and renewals extend from whichever is later — now, or your current expiry.
100 searches / month · self-serve key · no card
5,000 searches / month · everything in Free
30,000 searches / month · everything in Pro
Rummage already powers Research mode in production — the API is the same one the product uses.