Deploying is three steps, and only the first one takes any real time:
- Connect the Zywave MCP. One connector, added once, in the AI client you already use.
- Add a skill. From the skill’s page, in the client you picked.
- Paste the example prompt. Every catalog entry ships one that works as written.
About ten minutes the first time. Seconds per skill after that, because every skill reuses the same connector.
Start from what you want to do
The flow is personalised to the thing you are deploying, so start from a catalog entry rather than here:
- Skills catalog →
- Agents catalog → (placeholder entries)
Each entry has a Deploy button that walks you through the three steps for the client you choose. Every deploy flow on the site is also listed below, one per catalog entry.
Supported AI clients
| AI client | Support | Who sets it up |
|---|---|---|
| Claude | Works today | You, on your own |
| Microsoft Copilot | Works today, with an admin | An admin approves the connector once |
| ChatGPT | Works today, with an admin | An admin approves the connector once |
| Google Gemini Enterprise | Works today, with an admin | A Google Cloud admin sets up the connector once |
Custom integrations are possible too. Those talk to the MCP server directly, so start from the MCP Apps reference instead of here.
Before you start
- A Zywave account with access to the data you want to use.
- An AI client you already pay for: Claude, Microsoft Copilot, ChatGPT, or Google Gemini Enterprise.
- For Copilot, ChatGPT, and Gemini Enterprise, an admin who can approve a connector.
What is real here. The connection details on every step come from the Zywave MCP Server documentation. The click-by-click steps inside each client are written from those facts, not from screenshots, so menu labels may not match your version exactly. Nothing on this site installs anything for you, since there is no backend, so “add a skill” means pasting instructions today.