Automation and AI
MCP Servers & Tools
How to register external toolboxes (MCP Servers) so your Agents can take real-world actions.
An MCP Server is an external toolbox that an Agent can pick up and use. Once registered, it gives your Agent the ability to take real-world actions — like looking up a contact, creating a calendar event, or posting a message — without any of those capabilities being hardcoded into GreenVee itself.
What “MCP” means
Section titled “What “MCP” means”MCP stands for Model Context Protocol — an open standard that lets AI assistants safely use external tools through a standardized interface.
In practical terms: instead of GreenVee having to build a custom integration for every possible tool, any tool that speaks the MCP protocol can be plugged in. Your Agents connect to those tools through GreenVee’s MCP Server registry.
You do not need to understand the protocol. You only need to know the URL of the MCP Server you want to connect.
MCP Server fields
Section titled “MCP Server fields”| Field | What it does |
|---|---|
| Name | A label for this toolbox (e.g. “CRM Connector”, “Calendar Service”) |
| Endpoint | The web address GreenVee connects to in order to reach the server |
| Description | Filled in automatically from the server itself once connected |
| Status | The connection state — see below |
| Use in Polaris | An opt-in switch — when on, GreenVee’s shared Polaris assistant can use this server’s allowed Tools too |
MCP Tool
Section titled “MCP Tool”An MCP Tool is one specific action inside a toolbox. Tools are not created manually — they are discovered automatically when GreenVee connects to an MCP Server.
| Field | What it shows |
|---|---|
| Name | The action’s name (e.g. “look_up_contact”, “create_event”) |
| Description | What the tool does, in plain text |
| Permission | Whether an Agent may use this Tool freely — see below |
On each MCP Server’s detail page, tools are grouped into Read and Write (inferred from the action’s name) so you can see at a glance which ones only look things up versus which ones change data, plus an Other group for anything that doesn’t fit either.
Tool permissions
Section titled “Tool permissions”Every discovered Tool has a per-tool permission you control:
| Permission | Meaning |
|---|---|
| Allow | An Agent may call this Tool without asking you first. |
| Ask | An Agent must check with you before calling this Tool. |
| Deny | An Agent may never call this Tool. |
You can set permissions Tool-by-Tool, or set all Tools (or a whole Read/Write group) at once.
Connection status
Section titled “Connection status”| Status | Meaning |
|---|---|
| Checking connection | GreenVee is (re-)probing the server right now. |
| Connected | GreenVee can reach the server and its Tools are available to Agents. |
| Needs authentication | The server is reachable but requires you to sign in before its Tools can be used. |
| Error | The server couldn’t be reached, or returned an error — check the endpoint and try again. |
| Not yet connected | The server hasn’t synced yet (its default state right after you add it). |
Authentication
Section titled “Authentication”Most MCP Servers require you to authenticate before GreenVee can use their Tools. Each server’s detail page has an Authentication card with two options:
- OAuth 2.1 (recommended) — sign in with the server through a pop-up window. GreenVee keeps the connection alive afterwards without you having to sign in again.
- Static token — an escape hatch for servers that use a fixed API key or bearer token instead of OAuth. Paste the token once; GreenVee stores it securely and never displays it again.
You can disconnect and reconnect at any time, and re-check the connection on demand without waiting for the next automatic sync.
Each server’s detail page also lists every Agent currently using it, so you can see the blast radius before disconnecting or changing permissions.
Why this design exists
Section titled “Why this design exists”Hardcoding every possible tool into GreenVee would make the system rigid and slow to update. The MCP approach means:
- Extensible — any tool that supports the MCP standard can be plugged in
- Maintainable — tool updates happen on the tool’s server, not in GreenVee
- Controllable — you decide which Agents get access to which toolboxes
This lets you grow your Agents’ capabilities over time by simply registering new MCP Servers, without waiting for GreenVee to ship new built-in features.
Related pages
Section titled “Related pages”- AI Glossary — plain-English definitions for MCP Server and MCP Tool
- Agents — how to attach an MCP Server to an Agent
- Polaris — the shared assistant that can use servers marked “Use in Polaris”