> For the complete documentation index, see [llms.txt](https://docs.warp.dev/llms.txt).
> Markdown versions of each page are available by appending .md to any URL.

# Platform usage

Understand agent-hour charges for cloud runs and eligible local runs, including how platform usage differs from inference and compute.

Platform usage covers running and managing agents through Warp. Cloud runs, including [Warp Factories](https://docs.warp.dev/factories/) runs, and local runs on Business and Enterprise that use customer-supplied inference incur charges for billable agent time. Dollar-billed plans deduct platform charges from their usage balance; Enterprise billing follows your contract.

Customer-supplied inference is any setup where Warp isn’t paying for the model call: [Bring Your Own API Key (BYOK)](https://docs.warp.dev/agents/inference/bring-your-own-api-key/), a [ChatGPT subscription](https://docs.warp.dev/agents/inference/chatgpt-subscription/) for local Warp Agent conversations, [custom inference endpoints](https://docs.warp.dev/agents/inference/custom-inference-endpoint/), [BYOLLM](https://docs.warp.dev/enterprise/enterprise-features/bring-your-own-llm/), or [team-managed API keys and endpoints](https://docs.warp.dev/enterprise/enterprise-features/team-managed-keys-and-endpoints/) on Business and Enterprise plans.

Caution

BYOK and custom inference endpoint support are available for individual users and organizations with 10 or fewer employees, subject to Warp’s [Terms of Service](https://www.warp.dev/legal/terms-of-service). Companies or organizations with more than 10 employees need a Warp Business or Enterprise plan to use these features.

## How platform usage is measured

An *agent hour* is one hour of billable agent activity. Partial hours are billed proportionally. The standard platform rate on dollar-billed self-serve plans is $5 per agent hour; Enterprise rates follow your contract.

Agent hours are not the same as the elapsed run time shown in a dashboard. Time queued before the agent starts and idle time between turns while it waits for your next message are not billable.

Billable activity depends on the [harness](https://docs.warp.dev/platform/harnesses/):

-   **Warp Agent** - Time the agent spends generating a response. Command and tool execution time, such as a five-minute test suite, isn’t billable as platform usage.
-   **Claude Code and Codex** - Active runtime, including command execution and time spent waiting on the model provider.

## Inference, compute, and platform usage

These are cost components, not independent allowances. A run can incur more than one:

-   **Inference** - Model calls billed through Warp. Your own provider bills customer-supplied inference separately.
-   **Compute** - The sandbox on Warp-hosted compute. Local or self-hosted compute does not incur this charge.
-   **Platform** - Billable agent time for the runs listed below.

For example, a cloud run with Warp-provided inference on Warp-hosted compute incurs all three components. See [usage and billing](https://docs.warp.dev/support-and-community/plans-and-billing/credits/) for how they draw from your balance.

## When platform charges apply

Whether a run incurs platform charges depends on where it runs and who pays for inference.

### Uses platform services

-   **Cloud runs** - Billable agent time incurs platform charges regardless of the harness, inference provider, or compute host. Self-hosted workers avoid Warp-hosted compute charges, not platform charges. Enterprise exceptions follow your contract.
-   **Local runs on Business or Enterprise with customer-supplied inference** - Platform charges apply when the run uses BYOK, a connected subscription, a custom inference endpoint, or BYOLLM instead of Warp-provided inference.

### Doesn’t incur platform charges

-   **Local runs on Free, Build, or Max** - No platform charge, whether inference is provided by Warp or by you.
-   **Local runs on Business or Enterprise with Warp-provided inference** - Inference is charged without a separate platform charge.
-   **Regular terminal usage** - Shell commands and non-AI terminal features don’t consume agent usage.
-   **Third-party agent CLIs run directly** - Running `claude`, `codex`, or another CLI yourself, outside a Warp-managed run, doesn’t incur Warp platform charges.

## Tracking platform usage

In the Warp app, open **Settings** > **Billing and usage** to check account usage. Team admins can review the usage breakdown in the [Admin Panel](https://docs.warp.dev/knowledge-and-collaboration/admin-panel/) under **Billing**.

Detailed turn, conversation, and factory summaries show platform cost when available. These cost summaries are not agent-hour counters, and a run’s displayed duration is not a substitute for billable agent hours.

Platform charges draw from the billed user or agent’s eligible usage balances. See [included and purchased usage](https://docs.warp.dev/support-and-community/plans-and-billing/credits/#included-and-purchased-usage) for allocation and spending order.

## Related pages

-   [Usage and billing](https://docs.warp.dev/support-and-community/plans-and-billing/credits/) - How inference, compute, and platform costs are calculated.
-   [Purchasing additional usage](https://docs.warp.dev/support-and-community/plans-and-billing/add-on-credits/) - Buy usage or enable auto-reload.
-   [Bring Your Own API Key](https://docs.warp.dev/agents/inference/bring-your-own-api-key/) - Connect your own model provider keys.
-   [ChatGPT subscription](https://docs.warp.dev/agents/inference/chatgpt-subscription/) - Use eligible OpenAI models through your ChatGPT plan.
-   [Custom inference endpoint](https://docs.warp.dev/agents/inference/custom-inference-endpoint/) - Route Warp through any OpenAI-compatible endpoint.
-   [Plans, pricing, and refunds](https://docs.warp.dev/support-and-community/plans-and-billing/plans-pricing-refunds/) - Compare plans and refund policies.