--- name: olympus-setup description: Help the user install Olympus on their Mac and connect it to ChatGPT. Use when the user wants to set up or connect Olympus, when the Olympus dashboard says Olympus isn't on the Mac, or when Olympus tools report that Olympus is not connected. --- # Set up Olympus on a Mac Olympus answers from the user's own sources with an engine that runs on their Mac. ChatGPT reaches it through `https://mcp.olympusplugin.ai/mcp`, and the user approves ChatGPT on the Mac with one click. ## The one install command Olympus is installed with exactly this command, which the user runs themselves in Terminal on their Mac: ``` curl -fsSL https://olympusplugin.ai/install.sh | sh ``` It is also on https://olympusplugin.ai/install/. You never run it, or any other command, yourself: you show it to the user and they run it. Do not write, search for or suggest any other install command for Olympus: no `npm`, `bun`, `brew` or `git clone`, no other URL, and nothing from another site. Never suggest `sudo`, and never ask the user to edit files by hand. ## 1. Check what is already there Call `olympus_dashboard`. - `connection.state` `ready` or `installing`: Olympus is installed and connected. Show the dashboard and stop. - The call fails with "Your Mac is offline": Olympus is installed but the Mac is asleep or Olympus is not running. Ask the user to wake the Mac, and point them to https://olympusplugin.ai/help/mac-offline/. Then try again. - The dashboard says Olympus isn't on the Mac, or a tool says Olympus is not connected: ChatGPT is not connected to an Olympus engine yet. That is also what it says when Olympus is installed but not yet connected, so continue with step 2. ## 2. Install, then connect Ask whether Olympus is already installed on their Mac. - **It isn't, or they are not sure:** check that they have a Mac with Apple silicon (M1 or later) and macOS 13 Ventura or later. Intel Macs are not supported: say so plainly and stop. Otherwise tell them to open **Terminal** (in Applications, then Utilities), paste the install command above, and press Return. It needs no administrator password and installs for their macOS user only. It takes a minute or two and ends with "Olympus is installed and running". If it stops with a message instead, ask them to read it to you, explain it plainly, and point them to https://olympusplugin.ai/support/; do not suggest other commands. Running the same command again is safe. When it has finished, continue below. - **It is:** continue below. To connect, ChatGPT asks the user to connect Olympus (sign in) the first time a tool needs it. The page that opens must be in a browser on the Mac running Olympus: it asks to connect ChatGPT, and the user clicks **Approve**. That click on their own Mac is the only proof of ownership: there is no code to copy and no account to create. If they want to check that Olympus is running, they can run `olympus engine status` in a new Terminal window (it prints `"ok": true` when it is). ## 3. Confirm Call `olympus_dashboard` again. When it answers, tell the user Olympus is connected and point them to the Olympus dashboard in the ChatGPT sidebar to connect their first sources: Dropbox, Gmail or Google Drive, signing in with accounts they already have. Other sources the dashboard lists (such as X bookmarks, Readwise, Telegram or WhatsApp) are connected in Olympus on the Mac. Indexing starts on its own. Olympus downloads its built-in models in the background after install, so the first answers can take a while. ## Uninstall If the user asks how to remove Olympus, point them to https://olympusplugin.ai/install/#uninstall: one command in Terminal removes Olympus and keeps their data unless they delete it first. ## If something fails Report what the dashboard, the installer or the error says, in plain language, and stop. Do not retry with different commands or flags.