A hands-on workshop on architecting trustworthy agentic experiences in Slack using the Casey IT Support Agent.
- Join the Slack Developer Program
- Provision a Slack Developer Sandbox
- Install Slack CLI
- Obtain an Anthropic API Key
If you haven't already, join at api.slack.com/developer-program.
From the Developer Program dashboard, provision a new sandbox. This gives you a free Slack workspace with full API access for development. Use the Event Code provided in the workshop when provisioning the sandbox.
3. Install the Slack CLI
slack loginslack create agent support-agentFrom the selections that follow:
- Select the Support Agent option
- Select Bolt for JavaScript (as this workshop is run with JavaScript as the primary language)
- Select Claude Agent SDK as the LLM
cd support-agentWith the API key from your Anthropic dashboard, save the Anthropic API key to .env:
ANTHROPIC_API_KEY=YOUR_ANTHROPIC_API_KEYslack runYou should see Casey come online in your workspace (search for the app name - @support-agent - in the top-level search in the client). Send a DM to verify everything is working!
Each exercise corresponds to a layer of the Trust Pyramid. Complete them in order:
| # | Layer | Exercise | Directory |
|---|---|---|---|
| 1 | Safety | Build a Relevance Classifier | exercises/01-safety |
| 2 | Transparency | Add Task Cards to Workflows | exercises/02-transparency |
| 3 | Reliability | Structured State Management | exercises/03-reliability |
| 4 | Recoverability | Confirmation + Undo Flows | exercises/04-recoverability |
| 5 | Delight | Slash Command + Proactive Alerts | exercises/05-delight |
Each exercise directory contains:
- README.md — instructions and context
- Implementation files — reference code to study and port into the Casey app (e.g.
classifier.js,pto-request.js)
Raise your hand and one of our facilitators will come over!