Connect Your Employee to the World
In Teach Your Employee a Skill lesson, you gave your employee domain expertise through a custom skill. Now you will give it reach: the ability to communicate and act through the tools your profession actually uses.
Your AI employee can think, but it can only talk through WhatsApp. A real employee needs access to the platforms where your work happens. An accountant needs email for client invoices. A teacher needs Slack for parent communication. A developer needs GitHub for code review workflows. The connection you add here should be the single highest-value integration for your specific profession.
You have two paths: add a communication channel (Telegram, Gmail, Slack, Discord) using NanoClaw's built-in /add-* commands, or add an MCP server that gives your employee access to domain-specific tools and data sources.
The Challenge
Add the one connection that would create the most value for your AI employee's professional role. Then document exactly what data crosses the container boundary when that connection is active.
Acceptance Criteria
- Your new connection is working: demonstrated through a real send/receive interaction or a successful tool call
- You have documented what data crosses the container boundary (what goes in, what comes out, what is stored)
- You can explain in one paragraph why you chose this specific connection over alternatives
Deliverable
Add a section to your repo's README.md or create a connection-notes.md file documenting your choice, the data boundary analysis, and a screenshot or log showing the working connection.
Use Case Gallery
Pick the path that matches your profession, or adapt one to fit:
| Profession | Connection | Why This One | Path |
|---|---|---|---|
| Accountant | Gmail | Client invoice processing, payment status updates, receipt collection | Channel |
| Teacher | Slack | Parent-teacher communication, assignment reminders, classroom updates | Channel |
| Developer | GitHub MCP | Code review workflows, issue triage, PR summaries | MCP Server |
| Consultant | Google Calendar MCP | Meeting preparation, schedule analysis, client appointment management | MCP Server |
Your profession probably suggests a different connection entirely. That is the point: choose what matters for YOUR work.
Hints
PRIMM-AI+ Practice
This challenge follows the PRIMM-AI+ cycle. Before you build, predict. After you build, investigate.
Predict [AI-FREE]
Before you start this challenge, write down:
- Which connection (channel or MCP server) you will add and the one-paragraph justification you will write for your choice.
- What data will cross the container boundary in each direction (what goes out, what comes in, what is stored).
- What the first successful test interaction will look like: what you will send and what the employee will return.
- Your confidence score from 1 to 5.
Do not ask the agent until those notes are written.
Run
Use Claude Code ($ claude) to look up the setup steps for your chosen channel or MCP server. Then follow the NanoClaw setup flow (/add-telegram, /add-gmail, /add-slack, /add-discord, or MCP config) to complete the integration and test it with a real interaction.
Complete the challenge as described above: add the connection, demonstrate a working send/receive or tool call, and document the data boundary analysis.
Investigate
First, write your own explanation of whether the data boundary analysis matched your prediction (what actually crossed the container boundary versus what you expected). Then ask the agent: "Walk me through exactly what happens inside the container when I use this connection: what data is read, what is sent out, and what, if anything, is persisted?"
Modify
Change one requirement: add a second test that exercises a different capability of the same connection (for example, if you tested reading a calendar event, now test creating one; if you tested receiving a Slack message, now test sending one). Predict whether the employee's behavior and data exposure will differ, run the test, and verify.
Make [Mastery Gate]
Verify against the acceptance criteria above. Passing means: the new connection is working with a demonstrated real interaction, the data boundary analysis is documented, and you can explain in one paragraph why you chose this connection over alternatives.
Level 1: Where to Look
Run /help in your NanoClaw WhatsApp group to see all available commands, including the /add-* commands for channels. For MCP servers, check the NanoClaw repository's documentation on MCP configuration. Your Layer 3 design from the NanoClaw Hands-On Setup lesson already lists the MCP servers you planned: start there.
Level 2: Ask Your AI
Send this to your AI employee or Claude:
"Which communication channel or data source would give the highest ROI for a [your profession] assistant? Consider: frequency of use, time saved per interaction, and whether it enables tasks that were previously impossible."
Use the answer to confirm or adjust your choice before committing to the setup.
Level 3: Step-by-Step Guidance
For channels: Use the built-in commands: /add-telegram, /add-gmail, /add-slack, or /add-discord. Each command starts a guided setup flow. Follow the prompts, complete the authentication, and test with a simple message.
For MCP servers: Add the server configuration to your NanoClaw MCP settings. You will need the server's npm package or docker image, plus any required API keys. Test with a simple request that exercises the new connection, for example, ask your employee to read your latest calendar event or fetch a GitHub issue.
For the data boundary analysis: Ask yourself three questions: (1) What data does my employee send OUT through this connection? (2) What data comes IN from this connection? (3) Is any of this data stored inside the container?