MCP Servers
The Model Context Protocol (MCP) lets your agents connect to external tool servers and use those tools in live conversations. Add a server URL, complete any sign-in or API key step when asked, then choose which tools the agent may use.
What is MCP?
MCP is an open standard for connecting AI agents to external tools and data sources. Instead of building a custom integration for each system, you connect a server that publishes tools; the agent discovers them and can call them during a conversation.
Typical uses include checking appointments, updating records in your own systems, triggering workflows, and reading product or account data. Each tool carries its own description from the server, which is what guides the agent on when to use it.
Advanced feature
Adding an MCP Server
Follow these steps to connect an MCP server to your agent:
Open the agent editor
Open the MCP tab
Turn MCP on
Paste the server URL
Complete access if asked
Some servers need permission before tools can be listed:
- Sign in: Opens a short sign-in window for servers that use account access.
- API key: Paste a key or token when the server expects a static credential.
- Open access : If the server allows it, tools appear straight away with no extra step.
After access succeeds, review the tool list, enable the tools you want, then save the agent so the connection is stored with its configuration.
Sign-in and API keys
When a server requires access, Cloudax Connect prompts you before tools are listed:
- Sign in: Use this when the server asks you to approve access with an account. Finish in the pop-up window, then return to the editor.
- API key: Use this when your provider gave you a key or token. For the usual authorisation header, paste the token and Cloudax Connect adds the standard prefix for you.
Credentials are stored with the agent so later calls can use the same server. You can sign in again or replace a key from the server’s expanded panel if access expires.
Choosing tools
After a server connects, its tools appear in the editor:
- Enable or disable: Only enabled tools are available to the agent during conversations.
- Groups: Tools are grouped using information from the server so you can scan documentation, read and write tools more easily.
- Refresh: Pull the latest tool list when the provider updates the server.
When should the agent use a tool?
What you configure
Each connected server keeps these settings:
| Setting | Required | What it does |
|---|---|---|
| URL | Yes | The HTTPS address of the MCP server. Tools are loaded from this address. |
| Name | Yes | A short label shown in the editor. Often filled from the server when it provides one. |
| Active | No | When off, the agent keeps the connection saved but does not use that server’s tools. |
| Tools | No | Which tools from the server the agent may call. Turn individual tools on or off, or enable all. |
Managing servers
Connected servers appear in the MCP tab. From there you can:
- Expand: Open a server to see its tools grouped by type (for example documentation, read and write).
- Refresh: Reload tools from the server after the provider adds or changes them.
- Turn off: Use Active to pause a server without deleting it.
- Remove: Delete a server when you no longer need it. The agent loses those tools as soon as you save.
MCP settings are saved with the agent configuration, so changes follow the same version history as the rest of the agent.
Test before going live