For teams
Flat monthly rate, EU and UK data residency, no prompt training. No per-seat charge.
The Team plan provides up to five independent API keys, GPU and CPU model access, and the full workflow engine under a single flat monthly amount. Your prompts are never used for training. Your data does not leave your chosen AWS region as part of inference.
Commercial teams need predictable infrastructure costs and a clear data processing story. Marigold provides both: one fixed monthly amount and no third-party model provider in the inference path.
Account model
01
Up to five independent API keys on the Team plan. Assign one per service, environment, or team member. Each key is independently revocable. Pro tier has no key limit.
02
£79 per month. All keys share the same infrastructure and the same flat rate. No per-token billing, no per-seat charge, no usage alerts regardless of how the team distributes calls.
03
CPU and GPU models: instruct, embedding, TTS, image-to-text, depth, segmentation, and eval types. All available under one plan. See the full registry.
04
Inference runs in eu-west-2 (London) by default. No prompt reaches an external model provider. Data does not leave your chosen AWS region. See the architecture detail.
05
Inference inputs are not retained for training, fine-tuning, or any other purpose. This applies to all plan tiers. Pro tier can obtain a signed data processing agreement confirming this.
06
Build and register YAML workflow templates, submit executions, and query execution history. Parallel step execution, typed contracts, and a full audit trail. See the workflow docs.
Plan comparison
| Feature | Developer | Team | Pro |
|---|---|---|---|
| Monthly price | £19 | £79 | £299 |
| API keys | 1 | 5 | Unlimited |
| GPU models | No | Yes | Yes |
| Workflow engine | Yes | Yes | Yes |
| Eval surface | No | Yes | Yes |
| Custom model onboarding | No | No | Yes |
| Data processing agreement | No | No | Yes |
| Dedicated ECS capacity | No | No | Yes |
Frequently asked
Can we use Marigold for internal team tooling with sensitive data?
The architecture is designed for it. Inference runs within your chosen AWS region. No data is forwarded to an external model provider. Prompts are not retained. For regulated data, the Pro tier provides a data processing agreement that sets out retention, deletion, and processing obligations formally.
Is there a per-seat charge?
No. The Team plan is a flat monthly rate regardless of the number of people using it, up to the API key limit of five. There is no per-seat charge. Pro tier has unlimited keys and no user limit.
How do multiple API keys work?
Each key is an independent credential tied to the same account. Keys can be assigned per service, per deployment environment, or per team member. Any key can be revoked without affecting the others. Usage is not segregated by key at the billing level; all keys share the flat monthly rate.
What happens if we exceed rate limits?
Requests that exceed concurrency limits queue via SQS and are processed in order. You do not lose requests; you may wait. Dedicated ECS capacity on the Pro tier removes shared concurrency constraints entirely.
Team plan access is in limited release. Mention your team size and use case when you join.
No spam. One email when team onboarding opens.
Noted. We will be in touch.