Citation API with real page numbers
Send a sentence. Get an academic source that actually supports it, the page the claim appears on, and a formatted citation. Refusals are free — if nothing supports the claim, you pay nothing.
Free tier: 100 resolved citations, no card.
Already have an account? Your keys and usage are under API and keys.
curl -X POST https://cytado.com/api/ext/find-source \
-H "Authorization: Bearer $CYTADO_KEY" \
-H "Content-Type: application/json" \
-d '{
"items": [{
"id": "c1",
"claim": "Children aged 2-5 spend on average more than two hours a day in front of a screen.",
"hint": "screen time in preschool children"
}],
"langs": ["en"]
}'{
"results": [{
"id": "c1",
"found": true,
"source": {
"title": "…",
"authors": "…",
"year": 2021,
"page_from": 47,
"page_from_label": "47",
"snippet": "…",
"relevance": 9
}
}],
"rozliczenie": { "cytowania": 1, "tematy": 0 }
}What makes it different from a search API
Crossref, OpenAlex and Semantic Scholar are free and excellent at metadata — and none of them can tell you which page your claim is on, because they never open the document. cytado reads the paginated text, finds the passage, and reports the printed page number: the one on the page itself, the one a supervisor checking your footnote will land on. Every result also carries a link to the original, and for more than half of the corpus that link opens the full text — these are open-access publications.
Three endpoints
full documentation →POST /find-sourceA sentence in, a source with a page number out. The core call — this is what you are billed for.
POST /chapter-sourcesA chapter topic in, a set of citable sources out, ranked by how much of the topic they cover.
GET /usageWhat you have used this period, against your plan. The same numbers your invoice is built from.
Building an AI agent? Connect over MCP
The same API is exposed as an MCP server — your agent connects once and gets four tools whose descriptions explain when to call them and what is billed. No endpoint mapping, no prompt engineering about our mechanics: ask the agent to write with citations, and it grounds its claims itself. Works with Claude Code, the Claude API MCP connector, and any Streamable HTTP client. Same key, same billing, same usage dashboard as REST.
claude mcp add cytado --transport http https://cytado.com/api/mcp \
--header "Authorization: Bearer $CYTADO_KEY"Step-by-step guide (claude.ai, Claude Code, ChatGPT): connect the MCP connector; technical details in the documentation.
Pricing
| Plan | Price / mo | Citations | New topics | Extra citation | Extra topic |
|---|---|---|---|---|---|
| Test | free | 100 | 3 | — | — |
| Basic | $19 | 400 | 5 | $0.04 | $1.00 |
| Start | $49 | 1,200 | 15 | $0.04 | $1.00 |
| Pro | $99 | 3,000 | 40 | $0.04 | $1.00 |
A citation is one sentence resolved to a source with a page number. A new topic is building the corpus for a subject we do not have yet — it takes minutes of work and costs two orders of magnitude more, so it is metered separately. Overage is capped at an extra 100% of your plan — the bill can never surprise you by more than twice the plan.
What it will not do
It will not invent a source for a claim nothing supports — you get an explicit miss, not the nearest match dressed up as an answer. It will not attach a page number to a document without printed pagination. Sources we may not republish are returned as metadata and a short quotation under fair use, never as full text.