Developers / Quickstart
Your first projected call.
Three steps to the thing that matters: a response shaped to what your agent reads, not what
the upstream felt like sending. It takes about five minutes, and the last step is a single
parameter.
curl … --data 'select=id,doi,title,cited_by_count' - 01
Get a key in the console
Sign in and create a key — no card required to start. Your exact endpoint URLs are shown there too, next to the catalogue.
- 02
Connect over MCP, or reach for curl
Point an MCP agent at the endpoint and it discovers the catalogue as native tools — or make a plain REST call with your key as a bearer token. Same surface either way.
- 03
Ask for a section
Add a projection — the fields you actually read — and only those cross the wire. That single parameter is where the 76%-smaller number comes from.
- →
# claude_desktop_config.json — no API key changes hands
{ "mcpServers": {
"swarmsdb": { "url": "https://app.swarmsdb.ai/mcp" }
} } grant
{ "key_material": null,
"grant": "workspace-bound",
"revocable": true } curl https://app.swarmsdb.ai/data/openalex/works \
-H "Authorization: Bearer $SWARMSDB_KEY" \
-G --data 'search=graph neural networks' \
--data 'select=id,doi,title,cited_by_count' typed section
{ "shape": "table", "rows": 25,
"fields": ["id", "doi", "title", "cited_by_count"] } # one call, many parameter sets — results by signed webhook
POST /bulk/requests
{ "endpoint": "eurostat/data",
"params": ["nrg_bal_c", "demo_gind", "prc_hicp"],
"schedule": "0 6 * * *" } queued
{ "bulk_request_id": "brq_9f2c…",
"webhook": "hmac-signed" } upstream 8,457,367 → 1,508 chars
Point one agent at it.
Connect over MCP or REST. If your agents read more than one source, the context bill is the first thing you will see move.