# Try a starter nest.

> A starter nest is a run that Renest packed after it produced output, with every file checked on upload. Anyone with a Renest account can take one into their drive at no charge from the Starter nests page of the web console, then rent a GPU machine in their own RunPod or vast.ai account and paste the restore command the console gives them. Renest charges nothing for this; the machine is billed to you by the platform.

Canonical page: https://renest.ai/docs-starter-nests.html

<h2 id="what">A starter nest is a run we packed after it produced output</h2>

<p>We ran it on a GPU machine until it produced output, then packed the files and
dependencies behind that run into a nest. Every file was checked against its recorded
checksum when it was uploaded. We don't list a nest on the shelf until all of that is
finished.</p>

<p>Restoring a starter nest is the quickest way to see the whole path for yourself. You
don't need a working setup of your own, and you don't need a bucket or an access key. It is
the same path you would use for your own nests later.</p>

<h2 id="take">1 · Take one from the Starter nests page</h2>

<p>Sign in to the web console (a free account is enough) and open
<a href="https://app.renest.ai/trial">Starter nests</a>. Each row lists what you need to rent
before you spend anything:</p>

<ul>
  <li><strong>Packed on:</strong> the GPU we packed it on.</li>
  <li><strong>CUDA</strong> and <strong>Python:</strong> the versions its environment is
  locked to.</li>
  <li><strong>Rent at least:</strong> the disk to rent. The figure includes room for
  installing the dependencies, not just the nest's own files.</li>
</ul>

<p>Choose <strong>Take it</strong>. The nest goes into your drive as your own copy and uses
your drive's space. You get the same nest a hand-off would give you. If you take a nest you
already have, it uses no new space. If your drive doesn't have room, the console says how
much more you need and moves nothing. If the batch has been fully claimed, it says that too,
and new starter nests appear on the shelf.</p>

<h2 id="rent">2 · Rent a GPU machine in your own account</h2>

<p>Renest doesn't rent machines. The machine, its key and its bill stay in your own account
with the platform. After you take a nest, the console lists RunPod and vast.ai templates
that start from the <a href="docs-base-image.html">Renest base image</a>. That image
already has the tool installed and has nothing in it that would get in the nest's way. Any
Linux machine with an NVIDIA GPU works too. The template is just the least setup.</p>

<ul>
  <li><strong>GPU:</strong> the card in <em>Packed on</em> is a safe choice, but it doesn't
  have to match. Before anything downloads, the restore checks this machine's GPU, driver,
  memory and disk against what the nest needs. If the machine can't run the nest, it
  refuses up front.</li>
  <li><strong>Disk:</strong> at least the <em>Rent at least</em> figure, on the disk that
  will hold the folder you restore into. On RunPod that means the volume disk, under
  <code>/workspace</code>.</li>
</ul>

<p>Step-by-step pages for each platform:
<a href="docs-guide-runpod.html">Renest on RunPod</a> and
<a href="docs-guide-vast.html">Renest on vast.ai</a>. On vast.ai, the template's on-start
script is required. Without it the instance can't be reached.</p>

<h2 id="install">3 · Install or upgrade the tool on the machine</h2>

<p>Connect to the machine's terminal and run:</p>

<pre>$ uv tool install --upgrade renest
$ export PATH="$HOME/.local/bin:$PATH"
$ renest --version</pre>

<p>Run this on the base image too. The copy of <code>renest</code> built into the image is
older than the latest release, and <code>--upgrade</code> brings it up to date. On a
machine with no <code>uv</code> yet, install it first with
<code>curl -LsSf https://astral.sh/uv/install.sh | sh</code>. The version shown at the
bottom of this page is the latest release.</p>

<h2 id="code">4 · Issue a restore code from the nest's Restore page</h2>

<p>On the page you land on after taking the nest, choose <strong>Start the restore</strong>.
The Restore page walks you through five steps: <em>Confirm the version</em>, <em>Rent a GPU
pod</em>, <em>Install the Renest client</em>, <em>Issue a restore code</em> and <em>Run it
and check</em>. You can also reach it later from the nest's page in your drive.</p>

<p>At <em>Issue a restore code</em>, pick how long the code should last and issue it. The
code lets a machine download this one version of the nest until it expires.</p>

<ul>
  <li>It is shown once. If you lose it, issue a new one.</li>
  <li>It works on any machine until it expires. If a machine breaks, paste the same command
  on another one.</li>
  <li>Anyone holding it can use it, so keep it out of public places. You can revoke it from
  the same page at any time.</li>
</ul>

<h2 id="run">5 · Paste the command the console gives you</h2>

<p>The console gives you one block to copy. It writes the code to <code>grant.json</code>,
then runs the restore into <code>./run</code>. On RunPod, run <code>cd /workspace</code>
first so the restore lands on the volume disk:</p>

<pre>$ cd /workspace
$ cat &gt; grant.json &lt;&lt;'RENEST_GRANT'
{ …your restore code… }
RENEST_GRANT
$ renest restore --grant grant.json --dir ./run</pre>

<p>The restore goes through its stages in order. It checks this machine, downloads every
file and checks it against its recorded checksum, puts the files in place and installs the
dependencies from the lock. Then it starts the app and runs the packed workflow once as a
test render. For a fine-tuning nest, it runs the packed training once instead. It sends its
progress to your drive, so the <em>Run it and check</em> step in the console updates while
it works. <code>--no-report</code> turns that off.</p>

<p>The restore has done its job when that test run produces output, not just when the
command exits. If the connection drops or the machine stops partway, run the same command
again. Files already on disk that match their checksums are kept.</p>

<h2 id="next">Start the app yourself, then open it in your browser</h2>

<p>The restore starts the app only for its own check, then stops it. Its last lines,
<em>What's next</em>, say where your output is and which command starts the app. To start
it so that your own browser can reach it:</p>

<pre>$ renest start --dir ./run --listen 0.0.0.0</pre>

<p>Then open the app's port in your platform's panel. For ComfyUI that is 8188, unless the
start command names another port. <a href="docs-after-restore.html">After the restore</a>
covers the closing lines, reaching the app and copying your outputs off the machine.</p>

<h2 id="cost">What it costs: the machine is yours, the nest is free</h2>

<p>Taking a starter nest and restoring it costs nothing on Renest's side. It uses space in
your drive, and taking one you already have uses no more. The GPU machine is billed to you
by the platform, on the platform's terms. Renest never rents
machines and never sees your platform key. Copy off anything you want to keep before you
terminate the machine.</p>

<h2 id="stuck">If the restore stops, the last lines say why</h2>

<p>A refusal from the machine check means nothing has been downloaded yet. Rent a machine
that fits and paste the same command. For every other message, see
<a href="docs-troubleshooting.html">If a restore stops</a>. If you are still stuck, run
<code>renest support --dir ./run</code>, read what it prints and send it to us at
<a href="mailto:support@notify.renest.ai">support@notify.renest.ai</a>, or open a ticket
from the <strong>Support</strong> page in the console.</p>