Before you come

Do this before Friday 30 Oct. The venue Wi-Fi is shared by hundreds of people, so installs and downloads on the day are slow. It takes about 15 minutes, most of it the image download.

Join the workshop Slack

Questions before or during the breakout go to the GlassFlow Slack. Join it now, so you can ask for help on the day:

Scan the code, or open the Slack invite .

Install

1. A laptop with macOS or Linux. On Windows, use WSL2.

2. Python 3.11 or later. Check with python3 --version.

3. uv:

   curl -LsSf https://astral.sh/uv/install.sh | sh

   Check with uv --version: it prints uv 0.x.

4. Tares, version 1.51.1 or later:

   uv tool install tares

   Already have it? Upgrade with uv tool upgrade tares. On macOS you can use brew install glassflow/tap/tares instead.

   Check it:

   tares --version

   You should see tares 1.51.1 or later.

   If tares --version or, later, tares up fails with a traceback that mentions python3.10 (or older), uv installed Tares on a Python that is too old. Reinstall it on Python 3.12; uv downloads that Python for you:

   uv tool install --python 3.12 --reinstall tares

5. Optional: an agent or coding tool that speaks MCP (Claude Code, Cursor, or the agent you are building). In the last step of the guide it reads the same timeline as your Grok agents.

The demo and your key

1. Docker running (Docker Desktop or another Docker engine). Check with docker info --format '{{.ServerVersion}}': it prints a version number, not “Cannot connect to the Docker daemon”.

2. Download the demo stack and pull its images. This is the big download (about 600 MB of images):

   mkdir tares-demo && cd tares-demo
   curl -O https://raw.githubusercontent.com/glassflow/tares/main/demo/docker-compose.yml
   curl -O https://raw.githubusercontent.com/glassflow/tares/main/demo/inject.sh
   chmod +x inject.sh
   docker compose pull

   You should see api-server Pulled, prometheus Pulled and traffic-generator Pulled. Keep the tares-demo folder. You use it on the day.

3. An xAI API key, created at console.x.ai . Your event Grok credits work for this.

   • Why: the two agents you build call Grok, and the Tares running on your laptop needs a key to make those calls. The calls are billed to your xAI account.

   • When you use it: in step 3 of the guide, you paste the key into the Tares console on your laptop (Settings → Model providers). It stays on your laptop.

   • Keep it private: don’t share it or paste it anywhere else.

   • Check it now:

     export XAI_API_KEY=xai-...   # your key
     curl -s https://api.x.ai/v1/models -H "Authorization: Bearer $XAI_API_KEY" | jq -r '.data[].id'

     You should see a list of Grok model names. null or an error means the key is wrong.

If something is off

• tares: command not found after installing: open a new terminal. If it is still missing, run uv tool update-shell and open a new terminal.
• tares --version shows an older version: run uv tool upgrade tares.
• tares up fails with No module named 'exceptiongroup', or any traceback mentioning python3.10: Tares is on a Python older than 3.11. Run uv tool install --python 3.12 --reinstall tares, then tares up again. Your data in ~/.tares is kept.
• Cannot connect to the Docker daemon: start Docker Desktop and wait until it says it is running.
• The curl to api.x.ai answers 401 or 403: the key is wrong or has no credit. Create a new one at console.x.ai .