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

# Finding capacity

> How a workflow checks availability and quota before it creates a server, and what it does when nothing is free

You may ask CostGraph for capacity, such as a 32 core bare metal server or
eight GPUs in Europe. A workflow checks that it can be created before it asks
you to approve anything.

## What the workflow checks

The workflow lists candidates from your connected providers, cheapest first.
For each one, it reads the provider's live availability for that plan and
region. It then reads your account's limits and what you already use.

The first candidate that passes both checks is created, under your goal's
approval mode. A workflow does not send a create that a check already showed
would fail.

Providers CostGraph supports but you have not connected appear as
alternatives, with a link to connect them.

## Quota increases

When a candidate is at your account's limit and the provider lets you ask for
more, the workflow first looks for a request you already made. It reuses a
pending or granted request instead of filing a new one.

An increase can take hours or days. The workflow compares it with the next
candidate that is free now, using the price difference and any wait the
provider states. It writes that comparison on the run before it picks. If you
named an exact plan, region, or provider, only those options count.

An increase is an ordinary change, so it follows your goal's approval mode.

## Wait and recheck

While an increase is pending, or before it gives up, the run waits and checks
again. It checks every 15 minutes for the first hour, then hourly. Each check
reads live availability and quota, so a limit you raise in the provider's
console counts.

As soon as a check passes, the run creates the capacity. A run waiting on a
quota increase stops after 24 hours, and one waiting for stock stops after an
hour. The wait is counted from the first time the run had to wait. When it
stops, it shows why each candidate failed.

To wait longer or shorter, set **Wait for capacity** on the goal, from one
hour to seven days. That wait then applies to both.

A run that stopped on capacity offers **Check again** on its page. It reruns
the checks and creates the capacity if they now pass.

## Playbooks

Each run records how it checked a provider in a playbook. Your later runs use
it straight away. Other organizations use it only after the CostGraph team
reviews it. A playbook holds steps only, not your resources, prices, or
credentials.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.