Early-access plans
Plans below are indicative for design partners and early-access customers. Final commercial terms depend on deployment boundary, retention, support, and provider setup.Evaluation
For the first Python agentsContact us- Early-access Python SDK
- Evaluation workspace on managed runtime
- Task identifiers and basic event signals where configured
- Proposed AIOp volume scoped during review
- Community and product-team feedback channel
Recommended Production
For teams moving agents out of pilotsUsage-based/ AIOp volume, scoped during review- Durable orchestration and recovery paths
- Requestable review inputs for production reviews
- Configurable provider routing policy
- Model inference billed by your providers
- Security and deployment review before scale-up
Enterprise
For regulated and multi-team portfoliosCustom- Deployment boundary and retention review
- Support for multiple projects, teams, and providers
- Security, privacy, and procurement review package
- Commercial terms aligned with rollout scope
What the proposed AIOp unit measures
AIOp is a proposed commercial unit for runtime work. The order form must define the final metric and included volume.Included
Agent orchestration, retries, state transitions, event emission, policy checks, and review records where configured.
Excluded
Model inference, vector databases, cloud infrastructure outside the managed service, and third-party tools you contract directly.
Why it matters
The useful cost is not the cheapest model call. It is the cost of a reliable result that can be operated, reviewed, and improved over time.
Cost model
Model pricing and capabilities change quickly. The production cost moves to routing, reliability, supervision, and lifecycle control.Bring your providers
Keep commercial control of model providers and switch them as capabilities, prices, and compliance requirements change.
Route by policy
Decide when a small model is enough, when a stronger model is justified, and when another automated attempt should stop for project review.
Scale a portfolio
Move from a few pilots to many production agents without building separate visibility, retries, decision gates, and review inputs for each one.
Buying notes
The price conversation should map to the production boundary.Good fit
Teams that need durable Python agents, several model providers, task identifiers, available operating signals, review inputs, and a shared operating model between engineering, security, and business teams.
Current boundary
Teams looking for a finished no-code agent builder, a marketplace of prebuilt assistants, or a turnkey packaged deployment without design-partner work.