API Referencev1

Swarm API

Deploy parallel AI research agents, chain them into multi-step pipelines, and get back verified, source-cited outputs — all from a single assigned endpoint.

First result in ~30 s
2–20 parallel workers
Source-verified output
No SDK required

What do you want to do?

1

Get your endpoint

Log in → copy your assigned /public/<id>/ path

2

Deploy a swarm

POST /swarm/deploy/ with your query + swarmSize

3

Poll for results

POST /swarm/status/ every few seconds

4

Read your output

Worker output is returned inline in /swarm/status/ when allDone: true

Authentication

All requests use a public path as your credential. No bearer tokens — just set Content-Type: application/json and POST to your assigned path.

Base URLhttps://api.sushidata.ai/public/<your-assigned-id>/

Your first swarm in 3 requests

Copy, replace <id> with your assigned ID, run all three commands.

bash
1# 1. Deploy a swarm (takes ~5s)2curl -X POST 'https://api.sushidata.ai/public/<id>/swarm/deploy/' \3  -H 'Content-Type: application/json' \4  -d '{"query":"Research top 5 AI analytics competitors","swarmSize":5}'56# 2. Poll until done7curl -X POST 'https://api.sushidata.ai/public/<id>/swarm/status/' \8  -H 'Content-Type: application/json' \9  -d '{"workers":["do-7f3a2b1c","do-8e4b3c2d","do-9d5c4e3f","do-0c6d5f4a","do-1b7e6a5b"]}'1011# Worker outputs are returned inline in /swarm/status/ when allDone: true

✅ Quick Start complete — you just ran your first Swarm research job.

Typical runtime

30–90 s

swarmSize 5

Max swarm size

20 workers

per deploy

Max steps

10 steps

per sequence

Step timeout

10 min

auto-marked timed out