Background Agents

Background agents are Odin’s way of handling tasks that take more than a quick answer. When you ask Odin to do something that involves multiple steps — writing code, researching a topic, creating files, building a plan — it can spin up a headless agent to work through the task while you continue with whatever you were doing.

How to trigger an agent

Just ask naturally. Odin recognizes when a task is agent-worthy and launches one automatically:

“Research the best Rust web frameworks and save a summary to my desktop.”

“Write a Python script that watches my Downloads folder and moves PDFs to ~/Documents/PDFs.”

“Look up the Claude API pricing and compare it to OpenAI.”

When Odin decides to launch an agent, it responds with a short spoken acknowledgement (“on it”, “kicking that off”, “got it”) and the agent appears in the Agent Dock — a vertical stack of orbs on the right edge of your screen.

Note

Odin’s spoken response when launching an agent is intentionally brief. The agent’s progress is visible in the dock — Odin won’t narrate the steps it’s taking.

The Agent Dock

The Agent Dock is displayed as an overlay on the right edge of your screen. Each running agent shows:

• The task description
• Live output as the agent works
• Status (running / complete / failed)
• Cost

You can open multiple agents simultaneously. They run in parallel.

What agents can do

Background agents have access to the same tools as Odin’s main process:

• File system — read, write, and create files
• Web search — look things up
• Code execution — run shell commands, scripts
• MCP tools — any MCP servers you’ve connected
• Memory — read and write to Odin’s memory store

Stopping an agent

Click the stop button next to a running agent in the dock to cancel it. Partial work may have already been done (e.g., files created).

Agent history

Completed agents and their outputs are accessible in the conversation history. Agents that produce files will show the file path in their output.