Simple pricing, built for usage
Start free. Pay for what you deploy, not for seats. All tiers include the full agent runtime — tool calling, memory, guardrails, and tracing.
For prototyping and evaluation
- 1 active agent
- 1,000 invocations / mo
- Default guardrail policies
- Community support
For small production workloads
- 3 active agents
- 15,000 invocations / mo
- Custom guardrail policies
- Email support
For teams running agents in production
- 10 active agents
- 100,000 invocations / mo
- Webhook events
- Priority email + Slack support
Custom volume pricing, VPC deployment, and dedicated SLAs
Unlimited agents, private endpoints inside your VPC, custom data-retention terms, and a named support contact.
Pricing FAQ
One invocation is one call to agent.invoke() — a single turn of a conversation or task, including any tool calls the agent makes to fulfill it.
We'll email you when you cross 80% of your monthly limit. Overage is billed per-invocation at your plan's rate rather than cutting off traffic, unless you ask us to hard-cap it.
All plans can use any Bedrock foundation model your AWS account has access to. Model inference cost is passed through separately based on AWS's Bedrock pricing for the model you choose.
Enterprise plans support deployment into a customer-owned VPC. Reach out and we'll walk through the architecture and requirements.
Yes — tell us about your team when you request access and we'll factor that into your plan.