AI Glossary

This is the most important page in the Automation & AI section. Before you configure anything, read these definitions. Every term here maps directly to something you will see in GreenVee.

---

Agent

Plain English: A virtual worker you set up to handle a specific job.

An Agent is an AI assistant configured for one purpose — for example, “answer questions about open projects” or “triage incoming support requests”. You give it a name, a set of instructions, and decide which Skills and Tools it can use.

An Agent is not a generic chatbot. Each Agent has a defined role and only does what you have configured it to do.

Relates to: Agent Skills (what it knows), MCP Servers (what it can do), the Model (how it thinks).

---

Agent Skill

Plain English: A piece of know-how you teach an Agent.

A Skill is a named capability with its own instructions — for example, “Summarize meeting notes” or “Triage incoming requests”. You write those instructions once, then attach the same Skill to as many Agents as you like.

Think of it as training material. The Agent reads the Skill’s instructions and knows how to apply that capability when needed.

Relates to: Agents (a Skill can be shared across many Agents).

Caution

GreenVee has two different objects both called “Skill”. Agent Skills (this page) belong to the AI layer and are attached to Agents. HR Skills belong to the People module and are attached to Employees and Project Members — see Skills (People module). They share a word but are completely separate — changing one does not affect the other.

---

MCP Server

Plain English: A toolbox an Agent can pick up and use.

MCP stands for Model Context Protocol — an open standard that lets AI assistants safely use external tools through a standardized interface. You do not need to understand the protocol itself; what matters is what it gives you.

When you register an MCP Server in GreenVee, you are pointing the system at an external service (for example, your CRM, your calendar, or a data lookup service). GreenVee connects to that server and discovers what tools it exposes.

Relates to: MCP Tools (the individual tools inside the toolbox), Agents (which pick up one or more toolboxes).

---

MCP Tool

Plain English: One specific action inside a toolbox.

Each MCP Tool is a single operation that an Agent can perform — for example, “look up a contact”, “create a calendar event”, or “post a message”. Tools come from MCP Servers. You do not create Tools manually; they are discovered automatically when GreenVee connects to an MCP Server.

Relates to: MCP Servers (a Tool always belongs to one Server).

---

Workflow Definition

Plain English: A visual recipe for combining steps.

A Workflow Definition is a graph of steps (nodes) and connections (edges) that you build in GreenVee’s workflow editor. Unlike an Agent (which figures things out on the fly), a Workflow Definition always runs the same steps in the same order — A, then B, then C.

Use Workflow Definitions when the process must be consistent and predictable every time.

Relates to: Workflow Deployment (the frozen version that actually runs), Workflow Execution (one run of the recipe).

---

Workflow Deployment

Plain English: A snapshot of your Workflow Definition, frozen so it can safely run.

You edit a Workflow Definition freely, but GreenVee never runs your live, in-progress edits directly. When you click Deploy, GreenVee checks the graph for problems and, if it passes, freezes a numbered snapshot (v1, v2, v3, …). That snapshot — the Workflow Deployment — is what actually executes. Editing the definition afterwards doesn’t change anything that’s already deployed; you deploy again to ship the new version.

Relates to: Workflow Definition (the editable source), Workflow Trigger (bound to one deployed version).

---

Workflow Trigger

Plain English: The one thing that starts a workflow running.

Every deployed Workflow Definition needs exactly one Trigger — a start condition. GreenVee supports four kinds: Manual (you click Run), Schedule (cron-style: every interval, daily, weekly, or monthly), Webhook (an external system calls a URL), and a data change trigger (a chosen table gets a create/update/delete). Triggers can carry an optional condition so they only fire when specific values match.

Relates to: Workflow Deployment (one trigger per deployed version), Workflow Execution (a trigger firing starts one).

---

Workflow Execution

Plain English: A record of one time the recipe was followed.

Each time a Workflow Definition runs, GreenVee creates a Workflow Execution — a live record of what happened, when, and what the outcome was. You can watch it progress step by step on the Executions page, including any subagents or child runs it spawned.

Relates to: Workflow Trigger (what started the run), Workflow Definition (the recipe that ran).

---

Polaris

Plain English: GreenVee’s built-in, shared AI chat assistant.

Polaris is always available — there’s nothing to configure or activate. It’s shared across your whole organization and can use any Agent Skill or MCP Server your team has explicitly marked “Use in Polaris”. You can also start a Polaris chat bound to one specific Agent instead of the shared default.

Relates to: Agent Skills and MCP Servers (opt in via their Polaris flag), Agents (can be chatted with directly).

---

LLM / Model

Plain English: The brain inside the Agent.

LLM stands for Large Language Model — the AI engine that powers an Agent’s ability to read, reason, and respond. You may have heard of products like ChatGPT or Claude; those are built on LLMs.

In GreenVee, when you configure an Agent you choose which Model it uses from a curated set of about seven (including several Claude models, GLM, and MiniMax). Different models have different strengths, costs, and response speeds. The model is just the engine — the Agent’s instructions and Skills shape how it behaves.

Relates to: Agents (each Agent uses one Model).

---

How they fit together

Agent  ──uses──▶  Model (the brain)
  │
  ├──has──▶  Agent Skills  (know-how)
  │
  └──picks up──▶  MCP Server  (toolbox)
                      │
                      └──contains──▶  MCP Tools  (individual actions)

Polaris  ──may use, if opted in──▶  Agent Skills / MCP Servers

Workflow Definition  ──deploy──▶  Workflow Deployment  ──needs──▶  Workflow Trigger
                                        │
                                        └──fires──▶  Workflow Execution  (one run record)