Pay for scale, GPU execution, and persistent named graphs. The core wedge — declarative computation graphs with strict validation — ships in the free tier.
Everything you need to compose and run real DAGs.
Production scale with GPU execution.
For teams shipping multi-agent pipelines.
One DAG submission counts as one call, regardless of how many ops it contains (up to the per-call limit). Static validation failures do not consume the quota.
The Free tier executes on a CPU backend compiled through LLVM. The Pro tier compiles the same DAG to PTX for NVIDIA GPUs. Op semantics and validation are identical across backends.
Yes. The runtime, validator, and op catalog are open source. Paid tiers cover the hosted execution, GPU capacity, and workspace tooling.
For private deployments, regulated environments, or volume above the Pro tier, the Team plan is negotiated annually with an SLA.