Developers

The research tools on this site can be called from your own code or by your AI assistant: a REST API with one header, an OpenAPI document, and an MCP server. Any account can create an API key.

Quickstart

A working call in one minute: copy-paste curl, Python and JavaScript. Two of the three calls use no credits.

API reference

Every endpoint with its body, scope, price in credits and rate limit, and the errors with their meaning.

MCP server

The same tools inside Claude, Cursor, VS Code or Cline, installed with one command or one click.

Try it

The prior-publication check uses no credits. With a key from Settings, then API keys in SCIENCE_AI_API_KEY:

curl

curl -sS -X POST https://scienceaijournal.com/api/prior-publication \
  -H "Authorization: Bearer $SCIENCE_AI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"title":"Attention Is All You Need"}'

In Claude Code, the MCP server is one command (in other clients, see the MCP page):

Claude Code

claude mcp add science-ai-journal --env SCIENCE_AI_API_KEY=saij_YOUR_KEY -- npx -y science-ai-mcp-server

What you can call

Tool	Endpoint	Cost	Status
AI review (one agent per call)	POST /api/review/claude	4 credits per call	live
Prior-publication check	POST /api/prior-publication	no credits; limited per account to 10 calls a minute and 60 an hour (not the per-key limits); a key without tools:prior-pub counts as anonymous: 5 a minute and 20 an hour per IP address	live
Literature search and research-gap context	POST /api/research-gaps	no credits for the search	live
Journal Finder	POST /api/journal-recommender	15 credits per search	live
Pre-Check	POST /api/scorer	15 credits per call, once open	not open yet

Through the MCP server (science-ai-mcp-server on npm) an assistant can call find_research_gaps, check_duplicate_publication and ai_review_paper today. A refused call is never charged.

For AI agents

If you are an AI agent acting for a user: read the Markdown quickstart, ask the user for an API key (it starts with saij_), and send it as a Bearer token. Start with the prior-publication check or the literature search; they use no credits. Tell the user the price before a call that spends credits.

OpenAPI 3.1: every endpoint, body and error as JSON (https://scienceaijournal.com/developers/api/openapi.json)
quickstart.md: the quickstart as plain Markdown (https://scienceaijournal.com/developers/quickstart.md)
llms.txt: what the site and each tool does, for language models (https://scienceaijournal.com/llms.txt)
API catalog: the RFC 9727 link set that points at the three above (https://scienceaijournal.com/.well-known/api-catalog)

A journal, publisher or research organisation? Organisation keys carry exactly the tools you apply for: how the programme works.