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

# Compute quickstart

> Launch your first GPU or CPU instance—or a vLLM inference server—in minutes.

Get from zero to a running environment fast. This guide covers account setup, creating an instance, connecting over SSH, and launching a vLLM inference server.

> **You’ll need**
>
> * A **Compute** account (separate from a general Hivenet account)
> * Prepaid **credits** in your balance
> * An SSH client (macOS/Linux terminal or Windows Terminal)

<Steps>
  <Step title="Sign in to Compute">
    1. Go to **compute.hivenet.com** and select **Sign in**.
    2. Use Google, Apple, GitHub, or email/password.

    After sign-in you’ll land on the **Console**.

    <Tip>
      Add credits first to avoid interruptions. You can enable **auto top-up** later in Billing.
    </Tip>
  </Step>

  <Step title="Create a GPU or CPU instance">
    1. From the left sidebar, select **Create** (or **Create** from the Console card).
    2. Choose what you want to create: \*\*Container \*\*or **Virtual machine**.
    3. **Choose a location.** Pick what’s closest to your users or data.
    4. **Choose your setup.** Select **GPU** or **vCPU only**.
    5. **Pick a template or OS.** Examples include **Ubuntu 24.04 LTS**, **PyTorch 2.5**, or **Vulkan**. Templates and OSes come pre-installed, so you can start faster.
    6. **Connectivity.**
       * **SSH:** Paste your **public key**.
         * Need one?
           ```bash theme={null}
           ssh-keygen -t ed25519 -C "you@example.com"
           ```
           Then paste the contents of `~/.ssh/id_ed25519.pub`.
       * **HTTPS / TCP / UDP:** Keep defaults unless your app needs open ports.
    7. Review the summary and select \*\*\*\*Create instance \*\*\*\*or **Create virtual machine.**
    8. Wait for **Status: Running** on the **Instances** page.
  </Step>

  <Step title="Connect over SSH">
    From **Instances**, copy the SSH command shown for your instance, then connect:

    ```bash theme={null}
    ssh -i /path/to/private_key username@INSTANCE_IP
    ```

    <Note>
      \_username \_depends on the template. For Ubuntu images, it’s usually `ubuntu`.
    </Note>
  </Step>

  <Step title="Launch a vLLM inference server (optional)">
    If you want a ready inference endpoint:

    1. In the sidebar, select **Launch inference**.
    2. **Choose a model** to run (pre-trained options are listed in the UI).
    3. (Optional) **vLLM configuration.** Leave defaults for your first run. You can tune things later (e.g., context length, concurrency, sampling parameters).
    4. **Choose a location** and **setup** (GPU).
    5. Confirm **Connectivity** (SSH is required; HTTPS is available by default for the endpoint when applicable).
    6. Select **Run inference** and wait for **Running**.
  </Step>

  <Step title="Manage and clean up">
    * **Stop/Start:** You can stop and later start an instance from **Instances**.
    * **Terminate:** When you’re done, terminate the instance to stop billing.
    * **Billing:** Track spend and balance in **Billing**.

    <Warning>
      Billing is **per second while an instance is running**.

      **Data on the instance is deleted after termination.** Back up anything important before you terminate.
    </Warning>
  </Step>
</Steps>

## Troubleshooting

* **“Permission denied (publickey)”** when SSHing<br />Ensure you added the **public key** during creation, and you’re using the matching **private key** with correct file permissions (`chmod 600`).
* **Can’t reach the service**<br />Check that the required **port** is enabled under **Connectivity** and that any firewall rules are correct.
* **Instance won’t start or pauses**<br />Verify you have **enough credits**. Top up in **Billing**.

## Next steps

Now that you’ve launched your first instance, you’re ready to explore more:

* Create [custom templates](/documentation/essentials/custom-templates) tailored to your own workflows.
* Dive into [vLLM configuration](/documentation/essentials/how-to-vllm) to adjust context length, concurrency, and other parameters.
* Check [benchmark results](https://compute.hivenet.com/post/rtx-4090-and-5090s---up-to-2-3x-faster-than-an-a100) for RTX 4090 and RTX 5090 GPUs.
* Review [billing and credits](/documentation/account-billing/billing-info) to understand how per-second pricing works.
* Explore our [security overview](/documentation/reference/faq) for details on encryption and data handling.

## Need help?

* Join the Hivenet community on [Discord](https://discord.com/cTmPDPrMG2).
* Submit a request through our chatbot (you can find it on our [website](https://compute.hivenet.com)).
* Contact us anytime at [support@hivenet.com](mailto:support@hivenet.com).