Pricing

Ship ML pipelines and AI agents with confidence

Start open source and self-hosted. Upgrade to Pro for the managed control plane — on our SaaS or your own infrastructure.

One flat plan while we launch: $39 a month, every feature, no meters to read. The trial starts with full access and needs no credit card. Learn about Kitaru →

Open Source

For individuals and small teams

Free

Self-hosted · Your infrastructure, your limits


Includes

  • Import and record without limits
  • Cohorts, evaluators and experiment runs
  • Replay on your own workers, with your own keys
  • Community support
14-day free trial

Cloud

SaaS

For teams shipping agents to customers

$39/month

3 agents · 2 seats · 90-day session retention


Everything in Open Source, plus

  • Hosted dashboard and control plane
  • Replays and experiment runs included — no meters
  • Trial starts with full Enterprise access
  • Community support

Enterprise

SaaS

For organizations at scale

Custom

Unlimited agents · Custom seats and retention


Everything in Cloud, plus

  • SSO (SAML / OIDC)
  • Audit logs
  • Remote worker pools
  • Custom limits and retention

Compare every plan

One flat plan at launch — $39 a month, every feature, 14-day full-access trial. No meters, no usage math.

PlansOpen SourceCloudEnterprise
Plan
PriceFree$39 / monthCustom
Free trial14 days, full access
Credit card to startNeverNot for the trial
Replay & evals — every plan
Replay-based evals (Python & TypeScript)
Cohorts, evaluators and experiment runs
Replay on your own workers, with your own keys
Imports, recordings and scoring — no usage caps
Limits
AgentsSelf-hosted3Unlimited
SeatsSelf-hosted2Custom
Session retentionSelf-hosted90 daysCustom
Hosted control plane
Hosted dashboard and control plane
Role-based access (standard roles)
Enterprise & governance
SSO (SAML / OIDC)
Audit logs
Remote worker pools
Support
Support levelCommunityCommunityDedicated + SLA
View on GitHubSign up freeTalk to an engineer

Kitaru

Kitaru pricing, answered

What does Kitaru cost?
$39 a month, flat. The hosted Cloud plan includes 3 agents, 2 seats, 90-day session retention, and every feature. Replays, experiment runs, imports and recording all come with it. No meters, no usage math. Need more agents, seats or governance? That's the Enterprise plan.
What happens after the 14-day trial?
The trial is full access to everything and needs no credit card. When it ends, add a card and you're on the $39/month plan. Nothing about your workspace changes except the billing. There's no free hosted tier: the free option is the open-source version, self-hosted on your own infrastructure, forever.
What does the Enterprise plan add?
Unlimited agents, custom seats and session retention, plus the governance layer: SSO, role-based access and audit. If the Cloud plan's limits are the only thing in your way, that's the conversation to have with us.
How is this different from Langfuse, Braintrust or LangSmith?
They tell you what happened: traces you read, dashboards you check. Kitaru re-runs what happened. Your agent's real code executes again against the recorded world, so you can test your next change against your last thousand sessions before it ships. Kitaru also imports trace files exported from Langfuse, LangSmith, Braintrust, Logfire or Arize Phoenix, so your observability tool stays your system of record.
Can I self-host Kitaru?
Yes. Kitaru is open source under Apache 2.0, the same model as ZenML. One command brings up a local deployment in Docker, or self-host the server in your own VPC. Either way the worker runs beside your code, on your machine or your compute, using your own model keys, so replays never execute on our infrastructure. Cloud adds the hosted control plane and remote worker pools; Enterprise adds SSO, RBAC and audit.
Do you see my traces?
Self-hosted, no: the server and workers run in your infrastructure and your traces never have to leave your systems. On the hosted Cloud plan we manage the control plane, but replay still executes on your own workers with your own keys. The recorded sessions are yours.
Do I need a separate subscription for ZenML?
No, one subscription covers both workspaces. They're priced differently (ZenML's managed tiers scale with pipeline executions; Kitaru is a flat $39/month), but it's the same control plane, the same governance, and one bill.
What's included in Pro

Two products, one plan.

Switch SDKs without switching tools, billing, or governance. ZenML for reproducible ML. Kitaru for replay-based agent evals. Same control plane underneath.

ML pipelines

Pipelines & artifacts

Reproducible training, batch inference, evaluation. One DAG, versioned artifacts, every orchestrator.

  • Pipeline DAGs, artifact store, model registryVersion every step, every dataset, every model.
  • Run on Kubernetes, Vertex, SageMaker, AzureMLOne pipeline, any orchestrator — no rewrites.
  • Reproducible by default, replayable on demandRe-run any historical pipeline with one command.
Agent evals

Replay-based evals

Import your whole trace history, replay it against your next change, and keep every fix as a regression test — on one flat plan.

  • Sessions, cohorts, evaluators, experimentsThe whole replay loop — imports and scoring included.
  • One flat $39 plan — no meters to readReplays and experiment runs included at launch.
  • Seats for your domain expertsThe people who know what good looks like are not a line item.

Same control plane. Same governance. One bill covers both.

Are you startup or academic?

Apply for a special price to access ZenML Pro features for early-stage companies building ML-powered products, universities, research institutions, and educational use cases.

Apply Now

Trusted by 1,000s of members of top companies

Join the ZenML Community and start improving your MLOps

1,000,000

pipelines run in ZenML

100,000

pipelines run last month

21,000

stacks registered last 12 months

200,000

models trained last 12 months

"ZenML offers the capability to build end-to-end ML workflows that seamlessly integrate with various components of the ML stack, such as different providers, data stores, and orchestrators."

Harold Giménez

Harold Giménez

SVP R&D at HashiCorp

HashiCorp
AirbusAXABundeswehrEnelJetBrainsBrevoCross Screen MediaADEOLeroy MerlinMann+HummelKoble

Start deploying reproducible AI workflows today

Enterprise-grade MLOps platform trusted by thousands of companies in production.