This is a private and local agent built into the Seed desktop app.

To enable it, enter the desktop settings, click "Advanced" and set up an agent provider.

You may use an OpenAI subscription, or an API key from Anthropic, Google, or OpenAI. You can also connect the assitant to a self-hosted Ollama API.

This feature is implemented using Vercel AI SDK, so we can quickly add support for any of the AI SDK providers

The Agent Servers can run independently of the desktop app, so they are better suited for background tasks, respond to triggers, and to own certain responsibilities. Contrast that with the Assistant which runs locally, is private, and only responds on behalf of direct user instructions.

You can connect to our hosted offering, or self host your own agent server.

While creating your first agent you will be prompted to input a AI model provider. You can provide an API key from one of the major labs, or you may input an Ollama server.

The Agent Servers are implemented using the Pi SDK, so we can quickly add support for any of the Pi Providers

External agentic systems like Claude Code or Codex have the ability to download or "learn" skills. See more at agentskills.io to learn how skills work.

We distribute a skill for the Seed CLI in our skills repository, so your agents will know how the CLI can be used to read and write Seed content. To install the skill:

For the Seed Assistant or Agent Servers, you can connect to a custom Ollama server by providing the URL during setup.

You are responsible for configuring the security of your connection to the Ollama server. For example you may use a VPN to ensure that your desktop app can access your Ollama server, while blocking unauthorized access.