Field Guide

AI terminology, explained simply.

Understanding how AI is transforming our lives shouldn't need a PhD. Learn the latest terms in AI, and how you can put it into practice.

000 / 031

seen

Abilities

Abilities is the surface where the capabilities attached to a workspace or node live, connections and skills together.

Active work

Active work is the current set of responsibilities a party member, human or AI, is accountable for across everything they own.

Agent

An agent is a durable AI worker with its own identity and scope that owns real, ongoing work rather than just answering questions.

App

An app in Masterforce is a custom mini-application, your own tables, fields, and views, built without writing code.

Assistant

The assistant is the request-scoped chat helper that acts through the permissions of whoever invoked it, with no standing identity of its own.

Bring your own AI

Bring your own AI means running your own choice of model or provider, including entirely locally, instead of being locked into one vendor.

Checkout

Checkout is the lock an agent takes on something before it starts working, so two agents can't quietly overwrite each other's changes to the same thing.

Connection

A connection is a tool surface provided by an external integration, such as GitHub, Slack, or an HTTP API, that an agent or assistant can act through.

Context window

A context window is the amount of text an LLM can hold in a single call — everything it can "see" at once, including instructions, conversation history, and any documents provided.

Document

A document is an editable, markdown-based file, Masterforce's core format for writing, plans, and knowledge.

Fine-tuning

Fine-tuning is additional training on top of an existing model, using your own examples, so it permanently adjusts how the model responds.

Guardrails

Guardrails are the rules and checks that constrain what an AI is allowed to do, keeping it inside safe or intended boundaries even when a prompt or a model's own judgment might push past them.

Hallucination

A hallucination is when an AI states something confidently that isn't true — a fact, a citation, a piece of data — because it generated plausible-sounding text rather than checked a real source.

Harness

A harness is the runtime that gives an AI real tools, real permissions, and real memory instead of just a chat window.

Human-in-the-loop

Human-in-the-loop describes an AI system that pauses to ask a real person for a decision at specific points, instead of running fully on its own.

Integration

An integration is the specific way one of your apps uses a connection, for example, syncing Google Contacts into a people database.

Job

A job is one background execution run for a party member, a single attempt at moving a piece of active work forward.

LLM

An LLM (Large Language Model) is the underlying AI model — GPT, Claude, Gemini, and others — that actually generates text, reasoning, and tool calls for an agent or assistant.

MCP

MCP, short for Model Context Protocol, is the open protocol that lets an AI call real tools, like reading a file or updating a record, instead of only generating text about doing so.

Multi-agent system

A multi-agent system is an AI setup built from several specialized agents working together, instead of one general-purpose agent trying to do everything.

Node

A node is an element in the org tree under a workspace, identified by a slug path like co-x3/programs/knowledge, and it is the structural unit permissions and content attach to.

OKR

An OKR (Objective and Key Results) is a goal paired with the specific, measurable results that show whether it's actually being achieved.

Orchestrator

The orchestrator is the org-level coordinator that runs on a periodic heartbeat, noticing drift and nudging stalled work forward.

Party

The party is the combined view of the people and agents working together on the same node.

Party member

A party member is a standing actor in the workspace, a person or an agent, with its own identity, permissions, and ability to carry active work over time.

RAG

RAG (Retrieval-Augmented Generation) is a technique where an AI looks up relevant documents from a knowledge base and includes them in its prompt, instead of relying only on what it learned during training.

Skill

A skill is knowledge an AI reads on demand, a spec, template, or playbook it consults mid-task, rather than a standing tool it executes.

System prompt

A system prompt is the standing instructions an AI is given before a conversation starts — who it is, what it can do, and how it should behave — as opposed to whatever the user types.

Task

A task is a single unit of assignable work, owned by exactly one party member at a time.

Token

A token is the small chunk of text an LLM actually processes — roughly a word or part of a word — and it's the unit both context windows and AI usage costs are measured in.

Workspace

A workspace is the top-level container for one organization in Masterforce, holding its own structure, people, and settings.