Pricing
Plans
Every plan pays the same for tokens: $6.50 per million in and $18.50 per million out, on every model. A subscription changes what your key can reach and how fast, not what a token costs.
| What you get | Pro$19.99/mo | Premium$99.99/mo | |
|---|---|---|---|
| Endpoints | |||
Agent and swarm completions The standard endpoints, on every plan. | |||
Graph workflows /v1/graph-workflow/completions | |||
Agent batch /v1/agent/batch/completions | |||
Swarm batch /v1/swarm/batch/completions | |||
Reasoning agents /v1/reasoning-agent/completions | |||
Batched grid workflows /v1/batched-grid-workflow/completions | |||
| Models | |||
Full model catalogue Every model the platform lists, at the same token rate. | |||
Premium models GPT-5.6, Claude Opus 5, Gemini 3.1 Pro | |||
| Throughput and support | |||
Rate limits Requests per minute, and concurrent runs. | Base | Higher | Highest |
Compute lane | Standard | Priority | Priority |
Support | Community | Community | 24/7 priority |
| Billing | |||
Token rate Identical on every plan and every model. | Same | Same | Same |
Credit on sign-up | $5.00 | $5.00 | $5.00 |
Balance floor for premium endpoints Below this, a premium call is refused before any work runs. | — | $1.00 | $1.00 |
Monthly price | $0.00 | $19.99 | $99.99 |
New accounts start with $5.00 of credit on every plan. Usage is drawn from your balance; a subscription is billed separately from what you spend on tokens.
The premium endpoints
Five endpoints are subscriber-only. On a free key each answers 403 with an upgrade link rather than running, so nothing is charged for a request that was never going to work.
Graph workflows
POST /v1/graph-workflow/completionsDirected agent nodes and edges, compiled and run with parallel execution where the graph allows it. The backing endpoint for the Workflow Builder.
Agent batch
POST /v1/agent/batch/completionsHundreds of agent tasks in parallel in one request, for document analysis and other work that is wide rather than deep.
Swarm batch
POST /v1/swarm/batch/completionsThe same, for whole swarms: many collaborative runs executed concurrently instead of one after another.
Reasoning agents
POST /v1/reasoning-agent/completionsSelf-consistency, majority voting and iterative refinement, for answers that need to be right more than they need to be fast.
Batched grid workflows
POST /v1/batched-grid-workflow/completionsEvery task against every agent as a matrix, for comparing approaches side by side. The backing endpoint for the Grid.
Usage rates
Charged on top of the token rate, and identical on every plan.
Input tokens | $6.50 / 1M |
Output tokens | $18.50 / 1M |
Agent Per agent, on swarm and workflow endpoints. | $0.01 |
Image | $0.25 |
MCP tool call | $0.10 |
Exa search | $0.04 |
Web scrape | $0.15 |
Night discount Off token cost on swarm completions, overnight. | −50% |