Direct provider APIs
Pull usage data directly from provider APIs using credentials you supply.
Use direct provider APIs when the provider is the source of record for usage and your team can create a read-only credential for Doow.
Doow pulls usage data on a schedule from the provider API. No agent or middleware is involved.
When to use this path
Use direct provider APIs when the provider already exposes usage, billing, or metrics data and a provider admin can create a read-only credential. This is usually the lowest-effort usage-data path because Doow reads from the system that produced the usage.
Do not use this path when you need custom product metrics that the provider does not expose. Use the Instrumentation SDK for custom events, or Observability integrations when the metrics already live in an observability platform.
How it works
- A provider admin creates a read-only credential in the provider dashboard.
- You add the credential in Doow.
- Doow stores the credential encrypted and polls the provider's usage API on a regular schedule.
- Doow normalizes the usage data and attributes it to matching applications and licenses in your workspace.
Before you choose a provider
Confirm these items before opening a setup guide:
- which provider account, organization, project, region, or endpoint owns the usage
- who can create a read-only credential in that provider
- whether the provider exposes the usage fields Doow needs
- whether usage reporting has a normal delay
- whether cloud-hosted usage lives under a cloud platform path instead
Choose a provider type
Start with AI providers when the usage lives directly in an AI or API provider account. Start with cloud platforms when usage is exposed through a cloud monitoring, billing, account, subscription, or project layer.
Sub-categories
AI providers
Use these pages when usage lives directly in an AI or API provider account. Doow calls the provider's usage endpoint and reads provider-owned consumption data.
| Provider | What Doow reads |
|---|---|
| OpenAI | Token usage per model and project |
| Anthropic | Token usage per model including cache reads and cache writes |
| Groq | Input and output token rates per model |
| Fireworks AI | Prompt and cached token throughput per model |
| OpenRouter | Usage per model and routing endpoint |
| Replicate | Predictions and compute time per model |
| Authress | API call volume per resource and operation |
Cloud platforms
Use these pages when usage is surfaced through a cloud monitoring or billing layer rather than a direct provider usage endpoint.
| Provider | What Doow reads |
|---|---|
| AWS Bedrock | Token usage per model and region via CloudWatch |
| Azure OpenAI | Token usage per deployment and region via Azure Monitor |
| Google Vertex AI | Usage per model, project, and region via Cloud Billing |
Confirm the sync worked
After connecting a provider, open the integration detail page from Company Settings, then Integrations. A healthy sync shows a connected state, a recent sync timestamp, and usage rows for the expected organization, project, account, region, model, deployment, or endpoint.
Common blockers
Use these checks when credentials are accepted but usage does not appear:
- Confirm the credential belongs to the account, organization, project, region, or endpoint that owns the usage.
- Confirm the credential has read access to usage, billing, metrics, or organization-level data.
- Confirm the provider has usage in the selected date range.
- Check whether the provider reports usage with a delay.
- Check the integration event log for credential, scope, permission, or rate-limit errors.
Next steps
If you already know your provider, open its setup guide. If you are still choosing between usage-data paths, return to Choose a usage data source.