Skip to main content
Last updated on

Omnigent quickstart

Beta

Omnigent is in Beta. For the open-source documentation, see the Omnigent documentation.

Omnigent runs as a managed deployment on Databricks. The server that coordinates your sessions is managed for you, and you sign in with your Databricks workspace identity. This page covers the setup that is specific to Databricks. See the open-source Omnigent documentation for more information about Omnigent features and capabilities.

Prerequisites​

You must have:

To open Omnigent on a Databricks Sandbox, you must also have:

To install Omnigent using the CLI, you must have:

  • Python 3.12+
  • Node.js 22 LTS with npm and tmux

Choose a host​

Omnigent always runs on your managed Databricks deployment server. What differs is the host that executes your agent:

  • Databricks Sandbox (managed): Databricks provisions and runs the host in the cloud. It keeps running when your laptop is off, and your credentials never leave Databricks. Select this to get started quickly.
  • Your own machine (local): You install the Omnigent CLI and register your laptop or a VM as a host. Your laptop or VM becomes the execution environment so agents can access local files, tools, or networks. A local host is available only while your machine and the omni host process are running.

You sign in with your Databricks workspace identity either way, because both hosts run through the same managed server. Session visibility and sharing are the same for both: your sessions appear in the <workspace-url>/omnigent UI and can be shared with other workspace users regardless of which host runs them.

Launch Omnigent on a Databricks Sandbox​

The easiest way to run Omnigent is with a Databricks Sandbox: a managed cloud host that Databricks provisions for you. Your agent runs in an isolated, reproducible environment that stays on even when your laptop is asleep. Models route through your workspace's Foundation Model APIs over AI Gateway automatically, so there are no API keys to manage and your model usage inherits AI Gateway's governance and controls.

  1. In your web browser, go to <workspace-url>/omnigent.
  2. Select New session.
  3. Open the host picker and select Sandbox.
  4. Describe a task in the composer and send it to start the session.

The Omnigent workspace UI host picker with Databricks Sandbox selected.

Launch Omnigent on your own machine​

Install the Omnigent CLI and register your laptop or a VM as a host. The machine you register becomes the execution environment for your agents, so choose the machine that has access to the tools, files, and network your agent needs. A local host is available only while your machine and the omni host process are running.

  1. Install Omnigent with the databricks extra, which adds the managed-platform integration.

    Bash
    curl -fsSL https://omnigent.ai/install.sh | sh -s -- --extra "databricks"

  2. Set up model credentials. Run omni setup:

    Bash
    omni setup

    Running omni setup in the terminal to start adding harness credentials.

  3. Credentials are scoped per harness, so you must select a harness, then choose how to provide its credentials.

    • To invoke your workspace's Foundation Model APIs, choose Databricks and enter your workspace URL. This signs you into the workspace and routes that harness's model calls through Databricks Unity AI Gateway, so there are no API keys to manage and usage is governed and billed there.
    • Otherwise, provide your own API key, subscription, or gateway.

  4. Sign in with your Databricks identity and register the machine as an Omnigent host:

    Bash
    omni login <workspace-url>
    omni host --server <workspace-url>

  5. In your web browser, go to <workspace-url>/omnigent.

  6. Select New session.

  7. Open the host picker and select the machine you registered. The host you select determines the execution environment the agent runs in, so pick the host that has access to the tools, files, and network your agent needs.

    The Omnigent workspace UI host picker with a registered local host selected.

  8. Describe a task in the composer and send it to start the session.

Connect the desktop app or mobile app​

The Omnigent desktop app is optional and only available on macOS. Install the desktop app, then after you have created a host, select Connect to new server in the app and enter your workspace URL.

The Omnigent mobile app connects the same way. Install it from the App Store or use the mobile web UI in any phone browser, then enter your workspace URL. For install and setup details, see the Omnigent mobile docs.

If your workspace restricts network access, an admin may need to allow mobile devices to reach the workspace before you can connect. See Mobile network access.

On this page