SageOx is most useful as a habit, not a one-off. This guide walks the arc: connect the right surface, capture often, then put the context to work. Work through it in order.

1. Set up

Pick the setup path that matches how you actually work. Most people start in assistant-mode plus one of the recording surfaces. Code-mode with ox is for the subset of people whose AI coworker works inside a checked-out repo.

Repo-first setup with ox

If you work in a coding tool that starts from a local repo, connect the repo once. Everyone who joins later inherits the same context.

That is enough for normal onboarding. If the repo setup seems off, ox doctor is the troubleshooting step, not the day-one requirement. See Getting started for the full walkthrough.

Assistant-mode setup with MCP

If you are not working in a repo-first coding tool, start in assistant-mode:

• Use the MCP overview for the model and permissions.
• Use the connect wizard to connect your assistant.
• Use Getting the most out of SageOx via MCP to learn the prompt patterns and interactive cards.

Capture-mode setup

If your main job is to capture discussions rather than work in a coding repo, start with the recording surface that fits the moment:

• Use the Console Recorder in the browser today.
• Use Ox Dot for dedicated in-room capture.

2. Capture

The context is only as rich as what you feed it. Record the conversations where decisions actually happen.

• Record discussions — standups, design reviews, retros. Use the web Console Recorder or Ox Dot for in-room capture. Start with your first discussion.
• Import walkthroughs — paste a Loom, Figma, or Cap link. See video import.
• Narrate as you go — transcript quality drives extraction quality. Silence produces no context.

3. Use the context

Capturing is half the value. The payoff is when your team — and your AI coworkers — read it back.

If you use assistant-mode

This is the right default for ChatGPT, Claude Cowork, Claude Desktop, and similar desktop assistants with tool calling support.

• Ask explicitly for SageOx: "Search SageOx for what we decided about X."
• Use the interactive cards when they appear; they are faster than reading a paragraph and then acting somewhere else.
• Keep the MCP guide close until the prompting pattern becomes habit.

If you use code-mode

This is still the deeper path when your AI coworker works inside a checked-out repo.

After ox init, many coding tools run that automatically at session start. Re-run prime after clearing or compacting context. Ask question-shaped prompts in your editor — ox surfaces relevant prior decisions from your Ledger on its own.

If your desktop tool supports both

In tools like Claude Cowork, the best setup is often both, especially if you ask questions in the desktop app and code in local repos:

• Assistant-mode over MCP for live SageOx search, bubble access, and interactive cards inside the assistant.
• Code-mode with ox in the repos you actively code in, because it unlocks the deeper repo-side features like priming, code indexing, and richer coding workflows.

4. Advanced habits

Once the basics are routine, go further.

For everyone

• Cast a review to the room — put a recording or live meeting on the TV. See cast to a TV.
• Capture short decisions early — a two-minute voice note or hallway recap is often more valuable than waiting for a perfect formal meeting.

If you also code in repos

• Tend your Team Context — edit AGENTS.md and SOUL.md directly. See Team Context.
• Index your code — run ox code index so coworkers get semantic search across your repo.

What's next

• Record your first discussion — start the capture habit.
• Features overview — what the context becomes.
• Troubleshooting — when something doesn't go to plan.