> ## Documentation Index
> Fetch the complete documentation index at: https://intunedhq.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# How to optimize your Intuned cost
Intuned's cost is based on several usage metrics:
* **[Compute hours](/main/02-features/usage-pricing-monitoring-alerting#usage-metrics-reference)**—Time your machines spend executing APIs
* **[AI spend](/main/02-features/usage-pricing-monitoring-alerting#usage-metrics-reference)**—Dollars spent on AI processing
* **[JobRuns](/main/02-features/usage-pricing-monitoring-alerting#usage-metrics-reference)**—Number of JobRuns executed
* **[Machine size surcharge](/main/02-features/usage-pricing-monitoring-alerting#usage-metrics-reference)**—Additional cost for large/x-large machines
Visit [Usage and billing](/main/02-features/usage-pricing-monitoring-alerting) to learn how to monitor your consumption and set limits.
Check each project's breakdown on the [Usage page](https://app.intuned.io/usage) to identify which metrics to optimize.
Below are ideas for optimizing each metric.
## Compute hours
### Speed up individual runs
The faster your automation runs, the less compute time you use. Follow the techniques in [Make automations faster](/main/03-how-to/solve/make-automations-faster) to reduce execution time—including resource blocking, network interception, and batching evaluate calls.
### Reduce retries
When an automation fails repeatedly, it exhausts all retry attempts before failing. Each retry consumes compute time for a run that ultimately fails. Fix transient issues instead of relying on retries to work around them.
Use the [trace viewer](/main/02-features/observability-monitoring-logs#inspect-run-attempts) to debug failed runs and identify the root cause of errors.
### Combine related operations
Each API run has startup overhead—loading the browser, navigating to the site, and rendering assets. Reduce this by grouping related operations:
* **Pagination** — Instead of running one API per page, scrape multiple pages in a single run since the browser is already loaded.
* **Related tasks** — If you have several small APIs that hit the same site, consider merging them to avoid reloading the site for each one.
### Speed up AuthSession checks
If you have [AuthSessions](/main/02-features/auth-sessions) enabled, the check API runs before every API call to verify the session is still valid. A slow check API adds overhead to every run. Keep your check API fast by:
* Only checking for essential authentication indicators
* Using simple selectors that resolve quickly
* Avoiding unnecessary navigation or waiting
The techniques in [Make automations faster](/main/03-how-to/solve/make-automations-faster) also apply to your check API.
## AI spend
### Use cheaper models
Different AI models have different costs per token. When using [AI SDK methods](/automation-sdks/intuned-sdk/typescript/ai/functions/extractStructuredData), specify a cheaper model if your task doesn't require the most capable one.
### Use Intuned SDK methods with built-in caching
Intuned SDK methods like [`extractStructuredData`](/automation-sdks/intuned-sdk/typescript/ai/functions/extractStructuredData) have caching built in to avoid redundant LLM calls for identical inputs. Enable caching when processing similar pages.
### Use deterministic code when you can
If you're using AI for tasks that could be done with selectors and JavaScript, replace them. AI extraction is useful for unstructured or unpredictable content, but costs add up for predictable patterns. See [Replace AI code with deterministic code](/main/03-how-to/solve/make-automations-faster#replace-ai-code-with-deterministic-code).
## JobRuns
### Reduce Job frequency
Check how often your [Jobs](/main/02-features/jobs-batched-executions) run. If you're polling for data that rarely changes, reduce the schedule frequency.
### Skip unchanged items
If your Job processes multiple items but only some change between runs, skip items you've already scraped. Use the [KV cache](/main/01-learn/recipes/kv-cache) to track what's been processed and detect changes. This reduces the number of payloads processed per JobRun.
### Tune concurrency
When a Job runs, the concurrency setting in your [replication settings](/main/05-references/intuned-json#replication) controls how many machines run in parallel. More machines means faster completion but higher peak cost since you're charged for all active machines simultaneously. If speed isn't critical, reduce concurrency to spread the load over time.
## Machine size surcharge
Large and x-large machines cost more per hour. You typically need them for:
* JavaScript-heavy websites that consume lots of memory
* Automations that need to be very fast
* Automations with long-running browser sessions that accumulate memory usage
* Automations that require [headful mode](/main/02-features/stealth-mode-captcha-solving-proxies#headful-mode) or [stealth mode](/main/02-features/stealth-mode-captcha-solving-proxies#stealth-mode)
To reduce machine size requirements:
* Split long-running APIs into smaller chunks that process data in batches
* Follow the techniques in [Make automations faster](/main/03-how-to/solve/make-automations-faster) to reduce resource usage
* Disable [headful mode](/main/02-features/stealth-mode-captcha-solving-proxies#headful-mode) and [stealth mode](/main/02-features/stealth-mode-captcha-solving-proxies#stealth-mode) if not needed
## Upgrade your plan for better rates
Upgrading from the Developer plan to the Startup plan gives you lower overage rates:
| Metric | Developer | Startup |
| ------------------------- | ----------- | ----------- |
| Compute overage | \$0.12/hour | \$0.10/hour |
| JobRun overage | \$0.12/run | \$0.10/run |
| Large machine surcharge | \$0.18/hour | \$0.15/hour |
| X-Large machine surcharge | \$0.48/hour | \$0.40/hour |
If your monthly overages are significant, the Startup plan may pay for itself through reduced rates. See [Plans and billing](/main/05-references/plans-and-billing) for full pricing details.
## Need help optimizing?
If your costs are higher than expected, [reach out to us](/main/06-resources/help-and-support). We can review your usage patterns and suggest optimizations, or discuss custom plans for high-volume workloads.
## Related resources
Reduce execution time and compute costs
Monitor consumption and set limits
Plan features and pricing details