Customer Support

Create one on the API Keys page in this console. Send it on every request to https://api.swarms.world as an x-api-key header. The key identifies your account for billing and rate limiting, so treat it like a password and rotate it if it is ever exposed.

An agent completion runs one agent against your task and returns its output. A swarm completion runs several agents in a named architecture — the swarm type decides whether they run in sequence, in parallel, or under a director — and returns every agent output together with the aggregate. Use an agent when one model can do the job, and a swarm when the work splits into roles.

Every run is charged to your credit balance from the token usage the API actually reports, after the run finishes. New accounts start with $5 in free credit, and free credit is spent before credit you have paid for.

Plans are Pro at $19.99/mo and Premium at $99.99/mo. Subscribe from the pricing page. Once subscribed, the settings page shows your current plan and links through to Manage billing on swarms.world, where you can change plan, update a card, or cancel.

Requests are refused with HTTP 402 once your balance cannot cover them. Adding credit restores service immediately — there is no waiting period and no separate reactivation step.

Limits are per minute, per hour and per day, and they depend on your tier, so this console reads your live allowance from the API rather than publishing a table that would drift. The rate limits page shows the current numbers for your account.

A 429 means one of your tier allowances is exhausted. Retry with exponential backoff rather than immediately, and if you are hitting the limit steadily rather than in bursts, a higher tier raises it.

The catalog is served live by the API, and every model picker in this console is driven by it, so what you see here is what the API will accept rather than a list that can drift. The models page lists them all.

The API supports AgentRearrange, MixtureOfAgents, SequentialWorkflow, ConcurrentWorkflow, GroupChat, MultiAgentRouter, HierarchicalSwarm, MajorityVoting, CouncilAsAJudge, HeavySwarm, BatchedGridWorkflow, LLMCouncil, DebateWithJudge, RoundRobin, PlannerWorkerSwarm. Each one has a page describing what it is for, when it fits, and a runnable example.

The compare page runs the same task through several agent columns at once, each with its own model, system prompt and sampling settings, and shows the outputs side by side.

401 means the request carried no valid session or API key. 402 means your credit balance cannot cover the run. 422 means the request body failed validation — the response names the field. 429 means a rate limit. A 5xx is the platform or the upstream model provider, not your request.

Truncated output is usually max_tokens set below what the response needs. An empty output is usually max_loops of 1 on a task that needs a tool call and then a reply. The history page shows the token usage the API reported for the run, which tells you which of the two you hit.