Glossary
This glossary defines the terms you meet across AI Business OS, PRIMA Project Management, PRIMA Memory, PRIMA CRM and PRIMA Scholar. It also defines general terms for AI, AI agents and the tools the products run on, such as git, Node.js and Python. Where a page explains a term in full, the entry links to it.
A worker Claude starts in a separate conversation to handle a self-contained job, such as processing a long document, so the main conversation stays uncluttered. Agent definitions live in .claude/agents/.
Read more: document-processor
Agent Harness
Section titled “Agent Harness”The program that runs an AI model as an agent. It sends your request to the model, gives the model its tools, carries out the actions the model asks for and checks each one against your permissions. AI Business OS works through whichever harness you open in the workspace.
Read more: CLAUDE.md and AGENTS.md Files
AGENTS.md
Section titled “AGENTS.md”A pointer file that other AI coding tools, such as Codex, Cursor, Gemini CLI and Copilot, read in place of CLAUDE.md. It directs them to the same rules, and CLAUDE.md remains the main file.
Read more: CLAUDE.md and AGENTS.md Files
AI Agent
Section titled “AI Agent”An AI model given tools, so that it can read and write files, run commands and work through a task in several steps. Claude Code, Codex, Hermes Agent and OpenCode are AI agents. The Agent entry covers the helper agents defined in .claude/agents/.
Read more: Overview
AI Business OS Core
Section titled “AI Business OS Core”The base AI Business OS workspace that the installer puts in place first. Every PRIMA product except PRIMA Scholar is installed into it.
Read more: Installing AI Business OS
Anthropic
Section titled “Anthropic”The company that makes the Claude models and Claude Code. When you use Claude Code, the content Claude reads goes to Anthropic so the model can respond.
Short for application programming interface, the way one program asks another for data or for an action over the internet. PRIMA Scholar uses the APIs of academic databases, and an AI agent uses its provider’s API to reach the model.
API Key
Section titled “API Key”A secret string that identifies you to a service’s API. Keep API keys out of your workspace files and out of the conversation, because anything you paste into a conversation goes to the model provider.
Read more: API Keys and Configuration
Artificial Intelligence (AI)
Section titled “Artificial Intelligence (AI)”Computer systems that carry out tasks usually associated with human thinking, such as writing or recognising patterns. In these pages, AI usually means generative AI built on large language models.
Autonomy Levels
Section titled “Autonomy Levels”The rule that sets when Claude acts on its own and when it stops to ask you first.
Read more: Autonomy Levels
Blocked Commands
Section titled “Blocked Commands”Destructive actions, such as forced deletion or a force push, that Claude does not run without your explicit say-so.
Read more: Blocked Commands
Checkpoint
Section titled “Checkpoint”A file Claude writes to disk with a snapshot of the session’s progress, decisions, changed files and next action. /save writes one when you ask, PRIMA Project Management writes them automatically, and /resume reads them.
Read more: Resuming a Session
Claude
Section titled “Claude”The family of large language models made by Anthropic, which Claude Code runs.
Claude Code
Section titled “Claude Code”The AI agent harness from Anthropic, available in the terminal and as a VS Code extension. AI Business OS is built and tested with Claude Code, and these pages describe a Claude Code installation.
Read more: Requirements
CLAUDE.md
Section titled “CLAUDE.md”The file Claude reads automatically for instructions and context when it works in a folder. The main one sits at .claude/CLAUDE.md, and each domain and project can have its own.
Read more: CLAUDE.md and AGENTS.md Files
The AI agent harness from OpenAI, available in the terminal and as a VS Code extension. Codex reads AGENTS.md, and VS Code offers the Codex extension when you first open the workspace.
Read more: Configuration
Collection
Section titled “Collection”A named group of documents in your PRIMA Scholar research library. One document can belong to several collections.
Read more: Collections and Tags
Command
Section titled “Command”A saved set of instructions you run by typing a slash and its name, such as /day. Commands live in .claude/commands/.
Read more: Command Reference
Commit
Section titled “Commit”A saved snapshot of the changes in a git repository, with a message that describes them. /night commits your work at the end of a session, so git keeps the history of every change.
Compaction
Section titled “Compaction”What Claude Code does when a conversation grows too long for its working memory. It condenses earlier messages, so some detail from the start of the session is lost.
Read more: PRIMA Memory Overview
Company
Section titled “Company”A PRIMA CRM record for an organisation you deal with, with its relationship, status, contacts, engagements and interactions.
Read more: Companies
Company Context Files
Section titled “Company Context Files”The six files in .claude/company/ that describe your organisation, its team, audiences, voice, brand and industry. Skills read them so their output fits your business.
Read more: Company Context Files
Contact
Section titled “Contact”A PRIMA CRM record for a person, usually at a company, with their role, email, phone and how they first reached you.
Read more: Contacts
Contact Cadence
Section titled “Contact Cadence”How often PRIMA CRM expects contact with a company, in days. The defaults are 30 for clients, 60 for partners and 90 for suppliers, and you can change them.
Read more: Stale Clients and Contact Cadence
Context Window
Section titled “Context Window”The amount of conversation Claude can hold at once. When it fills, Claude Code condenses the conversation or the session ends.
Read more: PRIMA Memory Overview
Dashboard
Section titled “Dashboard”A single HTML page that PRIMA Project Management builds from your project records, with ten views and no server. /dashboard and /day build it.
Read more: Opening the Dashboard
Digest
Section titled “Digest”The dated report the PRIMA Scholar literature watch writes after each scan, listing the new papers it selected with APA7 references.
Read more: Literature Watch
Domain
Section titled “Domain”One of the seventeen numbered folders under Domains/ in a workspace created from AI Business OS 3.0 onwards. Each covers one business function, such as 02 Finance or 07 Marketing.
Read more: Domains
Embedding Model
Section titled “Embedding Model”A small model that turns a piece of text into a list of numbers that represents its meaning, so that texts with similar meanings sit close together. PRIMA Memory runs one on your computer for semantic search.
Read more: PRIMA Memory Overview
Engagement
Section titled “Engagement”A PRIMA CRM record of an arrangement with a company. Its type is sale, purchase, partnership or other, and only sales count in the pipeline.
Read more: Engagements
Engine Files
Section titled “Engine Files”The files AI Business OS owns and updates, such as its rules, skills, commands and scripts. .claude/engine-manifest.json lists them, and your own files are never on that list.
Read more: Updating Your Installation
Extract
Section titled “Extract”A summary file PRIMA Memory writes for a long session, kept in the extracts/ folder beside its index. A deep search reads these files.
Read more: How Indexing Works
Follow-Up
Section titled “Follow-Up”A reminder attached to a PRIMA CRM interaction, which the Today view in PRIMA Project Management’s /day shows when it falls due.
Read more: Interactions
Generative AI
Section titled “Generative AI”AI that produces new content, such as text, images or code, in response to a prompt.
Version control software that records every change to the files in a folder, so that you can see what changed and return to an earlier version. Every AI Business OS workspace is a git repository, and git also installs the workspace and delivers its updates.
Read more: Requirements
GitHub and GitLab
Section titled “GitHub and GitLab”Websites that host git repositories online. You do not need an account with either to install AI Business OS. You need one, or an account with a similar host, only if you want a remote repository.
Read more: A Remote Repository
Hallucination
Section titled “Hallucination”A statement from an AI model that sounds plausible but is wrong, such as an invented reference or figure. Check the facts, figures and sources an AI agent gives you before you rely on them.
Hermes Agent
Section titled “Hermes Agent”An open-source AI agent from Nous Research.
Homebrew
Section titled “Homebrew”A package manager for macOS that installs command-line tools with one command, for example brew install jq.
Read more: Requirements
A script Claude Code runs automatically at a set moment, for example before Claude edits a file or runs a command. AI Business OS uses hooks as safety checks.
Read more: Hooks Overview
Host Mode
Section titled “Host Mode”The conversation store PRIMA Memory reads. The installer sets Claude Code mode, so PRIMA Memory reads the Claude Code conversations for your workspace.
Read more: How Indexing Works
Interaction
Section titled “Interaction”A PRIMA CRM record of contact with a company or person, such as a call, an email, a meeting or a proposal, with a date and summary.
Read more: Interactions
A command-line tool that reads and writes JSON. AI Business OS and PRIMA Project Management use it to read and write the project register and the other state files.
Read more: Requirements
A plain-text format for structured data, written as names and values. The project register and most configuration files in the workspace are JSON files.
Read more: State Files and Validation
Keyword Search
Section titled “Keyword Search”A PRIMA Memory search that matches the words you give against your session files directly. It needs no index.
Read more: Searching Your History
Knowledge Cutoff
Section titled “Knowledge Cutoff”The date at which a model’s training data stops. A model knows nothing after that date unless an agent gives it a search tool or you supply the information in the conversation.
Large Language Model (LLM)
Section titled “Large Language Model (LLM)”An AI model trained on very large amounts of text to predict the next piece of text. Claude, the GPT models and DeepSeek are large language models. An AI agent uses one to decide what to write and which action to take next.
Literature Watch
Section titled “Literature Watch”A PRIMA Scholar routine that scans the academic databases for papers published on your standing topics since the last run.
Read more: Literature Watch
Local Model
Section titled “Local Model”A model that runs on your own computer, so that what it reads stays on your machine. Ollama is one common way to run one. A local model needs a capable computer and may be less capable than the largest hosted models.
Markdown
Section titled “Markdown”A plain-text format that marks headings, lists and links with simple symbols such as # and -. The instructions, rules, skills and most documents in the workspace are markdown files.
MCP Server
Section titled “MCP Server”A program that gives Claude extra tools through the Model Context Protocol. PRIMA Memory, PRIMA CRM and the two PRIMA Scholar servers are MCP servers.
Read more: How the Products Fit Together
Merge Base
Section titled “Merge Base”The copy of a file that an update compares against when it merges the new version with your edits. AI Business OS uses the published copy of the version your workspace runs.
Read more: Updating Your Installation
Milestone
Section titled “Milestone”A dated marker on a project, such as a launch or a review, that PRIMA Project Management records and the dashboard shows.
Read more: Pausing, Archiving and Milestones
The AI system that produces the responses, such as a Claude model. The agent harness gives the model its tools and carries out its actions.
Model Provider
Section titled “Model Provider”The company that runs a model and receives the content an agent sends to it, such as Anthropic for Claude or OpenAI for its GPT models. The provider’s own data policies apply to that content.
Model-Agnostic
Section titled “Model-Agnostic”Not tied to one AI model or provider. AI Business OS keeps its records and instructions in plain files, so that any capable AI agent can read and write them.
Read more: Overview
Node.js
Section titled “Node.js”A program that runs JavaScript outside a web browser. PRIMA Memory needs Node.js 22.5 or later, and PRIMA CRM, PRIMA Scholar and the PRIMA Project Management dashboard need Node.js 18 or later.
Read more: Requirements
The package manager that comes with Node.js. It installs the libraries that a Node.js program depends on.
Ollama
Section titled “Ollama”Free, open-source software that downloads open weight models and runs them on your own computer. VS Code offers the Ollama extension when you first open the workspace.
Read more: Configuration
Open Access
Section titled “Open Access”A paper anyone can read without a subscription. PRIMA Scholar marks every result as open access or not and lists open access papers first.
Read more: Searching the Literature
Open Weight Model
Section titled “Open Weight Model”A model whose maker publishes its trained weights, so that anyone can download it and run it on their own hardware. The Llama, Mistral, Qwen and DeepSeek families include open weight models.
OpenCode
Section titled “OpenCode”An open-source AI agent harness that runs in the terminal and works with models from many providers, including local models. OpenCode reads AGENTS.md.
Pipeline
Section titled “Pipeline”The open sale engagements in PRIMA CRM, grouped by stage, from lead and proposal through negotiation.
Read more: Engagements
Polite Pool
Section titled “Polite Pool”The service that OpenAlex and CrossRef give to callers who identify themselves with a contact email address. PRIMA Scholar joins it when you set OPENALEX_MAILTO or CROSSREF_MAILTO.
Read more: API Keys and Configuration
Project ID
Section titled “Project ID”The stable code that identifies a tracked project, such as P012. An ID is never reused, and new IDs follow the format already in your register.
Read more: Creating a Project
Project Link
Section titled “Project Link”A PRIMA CRM record that a company is involved in a project, with its relationship, department and notes. One company can link to many projects and one project to many companies.
Read more: Projects and Relations
Project Register
Section titled “Project Register”The file .claude/state/state.json, which lists every tracked project with its ID, status and details. The installer puts it in place, /setup fills in your details, and PRIMA CRM needs it to start.
Read more: State Files and Validation
Project Relation
Section titled “Project Relation”A PRIMA CRM record of how one project connects to another, as informs, supersedes, shares scope or part of.
Read more: Projects and Relations
Prompt
Section titled “Prompt”The instruction or question you give an AI agent. In AI Business OS, much of the prompting is already written into the rules, commands and skills, so a short request can carry a full method.
Prompt Injection
Section titled “Prompt Injection”Text in a document, web page or email that tries to give an AI agent instructions of its own. An agent that reads such text may act on it, so check what an agent proposes to do after it has read material from outside your workspace.
Push and Pull
Section titled “Push and Pull”The two git actions that move commits between your computer and a remote repository. A push sends your commits to the remote, and a pull brings commits from the remote to your computer.
Python
Section titled “Python”A programming language, which the processing-pdfs skill needs at version 3 together with the pypdf library.
Read more: Requirements
Reference Subfolder
Section titled “Reference Subfolder”A numbered folder inside a domain, such as 02.1 Accounts Payable, that holds standing material. It is never finished and is not tracked as a project.
Read more: Domains
Relationship
Section titled “Relationship”The standing a company has with you in PRIMA CRM, which is client, supplier, partner, internal or other. It sets the company’s default engagement type and contact cadence.
Read more: Companies
Remote Repository
Section titled “Remote Repository”A copy of your git repository on a hosting service such as GitHub or GitLab, or on another computer. It is optional. When you set one, /day, /sync and /night pull from it and push to it.
Read more: A Remote Repository
Repository
Section titled “Repository”The git copy of your workspace, usually held on a remote service such as GitHub so /sync and /night can push to it.
Read more: Projects, Workspaces and Repositories
Research Library
Section titled “Research Library”PRIMA Scholar’s local database of the papers and documents you import, with full-text search, collections and tags.
Read more: Your Research Library
Rollback
Section titled “Rollback”Undoing the last engine update by restoring the files saved before it. Type /update rollback or ask Claude to undo the update.
Read more: Updating Your Installation
A file in .claude/rules/ with a standing instruction that Claude loads in every conversation.
Read more: How Rules Work
Search Wizard
Section titled “Search Wizard”The PRIMA Scholar tool that reads your query, works out your discipline and suggests which databases to search.
Read more: Searching the Literature
Segment
Section titled “Segment”A saved PRIMA CRM filter over companies or contacts that you can run again at any time.
Read more: Segments
Semantic Search
Section titled “Semantic Search”A PRIMA Memory search by meaning, which finds sessions about a subject even when they used different words. It relies on an index that PRIMA Memory builds on your machine.
Read more: Searching Your History
Session
Section titled “Session”One continuous conversation with an AI agent, from the moment you open it to the moment you close it or start a new one.
Read more: Session History
Session History
Section titled “Session History”The list of recent sessions in your project records, each with a date, a summary and the projects worked on. /resume falls back to it when PRIMA Memory is not installed. The list holds the current calendar month, and earlier months move to a monthly archive in .claude/state/history/.
Read more: State Files Explained
Shared Project Log
Section titled “Shared Project Log”Entries in .claude/state/sessions/ in your workspace. PRIMA Memory reads them along with your own sessions when the folder exists, so colleagues who share the workspace can search each other’s entries. No product in the current releases writes to this folder.
Read more: How Indexing Works
A packaged workflow Claude follows for a kind of task, such as drafting documents or processing PDFs. Skills live in .claude/skills/.
Read more: creating-skills
Skill Pack
Section titled “Skill Pack”A set of related skills you install together. GenAI Skills Academy publishes skill packs in its learning community, including the six skills that left the template in AI Business OS 3.0.
Read more: AI Business OS Overview
SQLite
Section titled “SQLite”A database stored as a single file, with no server to run. PRIMA Memory keeps its search index in SQLite, and PRIMA CRM keeps its records in a SQLite database inside your workspace.
Staged Item
Section titled “Staged Item”Something PRIMA CRM’s sync has found that waits for you to accept or discard it. Examples are a message from an unknown person at a known company and a possible new lead. No email, calendar or meeting sync is built in the current release, so nothing is staged yet.
Read more: Email, Calendar and Meeting Sync
Stale Company
Section titled “Stale Company”An active PRIMA CRM company whose last interaction is older than its contact cadence. Leads and past companies are never stale.
Read more: Stale Clients and Contact Cadence
Stale Project
Section titled “Stale Project”A project in progress with no activity for more than seven days. After 28 days it moves to needs attention. In AI Business OS without PRIMA Project Management, /day and /status count day 7 as stale.
Read more: Status and Traffic Lights
stoppedAt and lastAction
Section titled “stoppedAt and lastAction”Two fields on each project. stoppedAt records where you stopped and what to do next, and lastAction records the last thing completed. /day and /resume read both.
Read more: State Files Explained
Task File
Section titled “Task File”An optional file at .claude/state/tasks/P###.json that holds a project’s tasks and their dependencies. The dashboard reads it when it exists.
Read more: Tasks and Dependencies
Template
Section titled “Template”A ready-made file or folder used as the starting point for a new project, client or document.
Read more: Templates
Terminal
Section titled “Terminal”The window where you type commands for your computer to run, such as git --version. VS Code has a terminal built in.
Three-Way Merge
Section titled “Three-Way Merge”How an update combines your edited copy of a file with the new version, using the merge base to tell your edits from upstream changes. Where both sides changed the same lines, the file is marked for you to resolve.
Read more: Updating Your Installation
Timeline
Section titled “Timeline”A Gantt-style chart of your projects that /timeline builds and opens in your browser, with its address copied so you can view it in VS Code’s Integrated Browser. Projects need a start date and a due date to appear.
Read more: The Timeline
Today View
Section titled “Today View”The dashboard view that /day opens, with health cards, where you left off, the next seven days of events and what needs you.
Read more: The Today View
The unit of text a model reads and writes, often a whole word or part of a longer word. Context windows, model limits and usage charges are counted in tokens.
An action an AI agent can take beyond writing text, such as reading a file or running a command. MCP servers add tools.
Tracked Project
Section titled “Tracked Project”A folder named {ID} {Project Name} that holds a piece of work with an objective and an end, registered in the project register. /newproject creates them.
Read more: Projects, Workspaces and Repositories
Traffic Light
Section titled “Traffic Light”The colour PRIMA Project Management gives each project in /day and /status. Green is active, yellow is stale (more than seven days without work), red needs attention (overdue, or 28 days or more without work), orange is paused and white is archived.
Read more: Status and Traffic Lights
Untracked Work Warning
Section titled “Untracked Work Warning”A dashboard warning on a project in progress or overdue that you have worked on in the last 14 days, when its tasks show no change in that time.
Read more: Tasks and Dependencies
VS Code
Section titled “VS Code”Visual Studio Code, the free code editor from Microsoft. You open the workspace in VS Code and work with your AI agent in its panel or in the terminal.
Read more: Requirements
Workspace
Section titled “Workspace”The folder you open in VS Code that holds your projects, configuration and .claude/ folder. A workspace is also a git repository.
Read more: Projects, Workspaces and Repositories
Workspace Identity
Section titled “Workspace Identity”The name, ID, type and repository recorded in state.json. PRIMA Project Management compares it with your git remote before every push and stops the push when they differ.
Read more: Rules It Installs
Windows Subsystem for Linux, which runs Linux inside Windows. AI Business OS may run on Windows through WSL, although Windows is not directly supported.
Read more: Requirements