Developers

Every fact, one request away.

Three ways in: free bulk files for whole datasets, a REST API for querying and lookups, and an MCP server so AI assistants can use AIDB as a tool. Everything returns sources and licenses.

Bulk files — free, no key

Every dataset is published as a static file in 9 formats, served from a global CDN. Append the extension to any dataset URL.

/d/{slug}.csv

CSV — Excel, Sheets, pandas, R

/d/{slug}.xlsx

Excel — Native Excel workbook

/d/{slug}.json

JSON — Array of objects

/d/{slug}.ndjson

NDJSON — One JSON object per line

/d/{slug}.tsv

TSV — Tab-separated

/d/{slug}.sql

SQL — CREATE TABLE + INSERT (SQLite/Postgres)

/d/{slug}.xml

XML — Generic XML rows

/d/{slug}.md

Markdown — Table for docs and LLM prompts

/d/{slug}.html

HTML — Standalone table page

curl -O https://aidb.si/d/exoplanets.csv

Full catalog as JSON: /api/v1/datasets. Entity pages export too: /e/Q937.json, .csv, .xlsx, .md, .jsonld.

REST API

Base URL https://aidb.si/api/v1. JSON by default; add format=csv (or any bulk format) to most endpoints. CORS is open. The machine-readable spec is at /openapi.json.

GET/api/v1/search?q={query}

Search entities and datasets. mode=prefix for autocomplete-style matching; limit up to 50.

GET/api/v1/entities/{id}

Every fact about one entity (Wikidata id like Q937): typed values, qualifiers (dates), sources, identifiers and time series. props=P569,P106 to filter; format=csv|xlsx|md….

GET/api/v1/entities?ids=Q1,Q2,…

Batch lookup of up to 50 entities in one call.

GET/api/v1/datasets

The catalog: every dataset with columns, license, sources and file URLs. Free and cached.

GET/api/v1/datasets/{slug}

Metadata and schema for one dataset.

GET/api/v1/datasets/{slug}/rows

Filter, sort, select and page rows — see the query language below. Responses include total and a next link.

GET/api/v1/categories

All topics and their datasets.

GET/api/v1/usage

Your plan and today's remaining allowance.

curl "https://aidb.si/api/v1/entities/Q312?props=P2139,P1128"

Query language

Filters are query parameters named after a column key, with an operator prefix. Combine as many as you like.

height_m=gte.400	greater than or equal (also gt, lt, lte)
country=eq.USA	equals (default if no operator)
status=neq.retired	not equal
name=like.tower	contains, case-insensitive
name=starts.new	starts with
country=in.USA,CHN,ARE	one of a list
architect=is.null	is null / is.notnull
select=name,height_m	choose columns
sort=-height_m,name	sort; - for descending
q=shanghai	free-text search across text columns
limit=100&offset=200	paging (max rows per request depends on plan)
format=csv	csv, tsv, xlsx, ndjson, sql, xml, md, html (default json)

MCP server — give your AI the database of everything

AIDB speaks the Model Context Protocol. Connect it once and your assistant can search, read entity facts and query datasets on its own — and cite AIDB pages in its answers. Tools: search, get_entity, list_datasets, describe_dataset, query_dataset.

https://aidb.si/mcp

Claude (web or desktop): Settings → Connectors → Add custom connector
Name: AIDB
URL:  https://aidb.si/mcp

For LLM crawlers and agents there's also /llms.txt — a plain-text map of everything here.

Limits & keys

No key needed to start: anonymous use gets 500 API requests and 25,000 rows a day, up to 1,000 rows per request. Bulk files never count. Paid plans raise every limit and come with an API key — send it as Authorization: Bearer <key>. Every response carries X-RateLimit-Remaining; when you hit the cap you get HTTP 429 until 00:00 UTC.

See plans