Resources, domains, plans, and costs
Register venture resources, connect domains, understand plan limits, and monitor AI spending.
Resources
Resources are the venture's infrastructure and service registry. Supported records include repositories, cloud services, databases, hosting, analytics, communication, and other operational assets.
Resource records can hold:
- provider and resource type;
- display name and description;
- URL or external identifier;
- non-secret configuration metadata;
- connection and health state;
- an optional credential relationship.
Use the secure Credentials area for tokens and secrets. Do not place raw credentials in descriptions, URLs, or general metadata.
Workflows receive an explicit resource selection at launch. Agents can then request time-limited, least-privilege access through the Security Officer flow without placing the source secret in their prompt. See Resources and Security Officer.
Domains
Domain tooling associates hostnames with ventures and deployment resources. Domain state can be inspected and reconciled by the platform.
Your DNS provider can still require an external record change or verification step. Follow the domain status shown in Organ and confirm the final record at the DNS provider.
Plans and limits
Plans control portfolio and usage entitlements such as venture count. The product can enforce configured limits when new ventures or restricted capabilities are requested.
Costs
Cost reporting separates AI-model usage from container compute:
| Cost | Source | Refresh behavior |
|---|---|---|
| Model | Recorded input, output, cache-write, and cache-read tokens multiplied by catalog pricing. | Finalized as phase/run usage is persisted; live token counters can precede final cost. |
| Compute | Completed runtime weighted by workload tier and capacity rate. | Settles through an hourly-refreshed attribution view. |
Venture Cost Analytics can break model spending down by date and workflow. Developer-mode detail adds major phase, status, failure cost, token, department, average-cost, and compute views.
The displayed model total includes recorded workflow runs and Brain synthesis. Compute remains a separate card and is not added into “Total Cost.”
Organizations can configure spending thresholds. They can restrict supported new work but do not cancel active runs, so use provider-side limits when a strict ceiling is required. Compute is an attribution estimate for completed runs, not a reconciled cloud invoice.
For formulas, prompt-cache accounting, billing meters, model selection, and diagnosis, see Compute, tokens, and model selection.