> ## Documentation Index
> Fetch the complete documentation index at: https://docs.b3os.org/llms.txt
> Use this file to discover all available pages before exploring further.

# Billing and Compute Units

Understand B3OS usage accounting, compute units, workflow pauses, and cost-aware workflow design.

B3OS tracks workflow usage with compute units (CU). CU accounting helps organizations understand automation cost, enforce limits, and pause workflows safely when usage constraints are reached.

## What Can Consume CU

| Activity                    | Examples                                                                    |
| --------------------------- | --------------------------------------------------------------------------- |
| Workflow runs               | Manual runs, scheduled runs, webhook runs, event-triggered runs             |
| Node executions             | Built-in logic, catalog actions, connector calls, wallet actions            |
| Caddie usage                | Workflow planning, repair, search, inspection, and debugging                |
| Provider work               | Action runtime calls that require external requests, chain data, or compute |
| Streaming and observability | Activity inspection and run debugging surfaces                              |

## CU-Aware Design

**Filter early**

Add branch conditions before expensive provider calls or wallet actions.
  **Batch where possible**

Prefer actions that fetch lists or aggregate data over many single-item calls.
  **Set clear schedules**

Avoid unnecessarily frequent cron schedules for low-priority checks.
  **Stop noisy failures**

Fix repeated provider or schema errors quickly so workflows do not burn usage on avoidable retries.

## Pauses and Limits

If usage constraints are reached, B3OS can pause workflows with a `cu_exhausted` reason. Review usage, adjust the workflow, or update billing before resuming.

**Pausing protects production workflows**

A paused workflow keeps its definition and history. Automatic trigger execution stops until the organization resolves
  the limit or resumes intentionally.

## Operational Checklist

| Check                         | Why it matters                                                       |
| ----------------------------- | -------------------------------------------------------------------- |
| Watch high-frequency triggers | Event-heavy triggers can create many runs quickly                    |
| Audit failed workflows        | Failures still consume execution resources                           |
| Review Caddie-heavy usage     | Caddie planning and debugging can be useful but should be deliberate |
| Monitor connector errors      | Expired provider credentials can cause repeated failed runs          |
| Track wallet workflows        | Transaction workflows may include reads, simulations, and writes     |