HQ Glossary

Acceptance criteria: The specific, checkable conditions that must be true before a task counts as done ("page loads in under 2 seconds", "test suite passes"). HQ's execution loops verify work against these instead of trusting "looks done". Defined per user story in a PRD.

Back pressure: The automatic checks (tests, linting, acceptance criteria) that push back on work before it's accepted. Back pressure is what keeps autonomous execution honest: the work isn't done until the checks pass.

Bot: An AI teammate you message like a person in the HQ app. A local bot runs on your own computer with your own Claude Code, Codex, or Grok sign-in. A cloud bot runs in your company's cloud and is always on (the HQ Console calls these Fleet Agents). See Local Bots and Cloud Bots.

Checkpoint: A quick mid-session save of your progress: summary, files touched, next steps. Run /checkpoint during long sessions. Lighter than a handoff. See Sessions, checkpoints, and handoffs.

Cloud-backed: A company connected to HQ Cloud, so its contents sync across machines and teammates. Companies are local-only until you cloud-back them with /designate-team. See HQ Cloud explained.

Command: A slash command you type to make HQ do something, like /prd or /hq-sync. Commands are implemented by skills. See Commands vs skills vs workers.

Company (tenant): A business or team you work with, held in its own sealed compartment under companies/. Each company's knowledge, credentials, and policies never mix with another's. See Companies, projects, and repos.

Deploy: Publishing an artifact (a report, page, or demo) to a private shareable web link with /deploy. See Sharing your work.

DM: A direct message sent to a teammate's HQ app with /dm. It shows up under Direct messages in the HQ window. Can carry a ready-to-run prompt or be scheduled for later. You can message people outside your company too, by searching for them in the HQ app.

Handoff: The end-of-session package written by /handoff: a summary, next steps, files touched, and git state, saved so a fresh session can pick up exactly where you left off. Stored under workspace/threads/. See Sessions, checkpoints, and handoffs.

HQ: Your AI operating system, one folder that organizes everything an AI assistant needs to work with you and your team (knowledge, projects, code, and rules) plus the commands to run that work. See What is HQ?.

HQ Workforce: HQ's paid team plan ($500/mo flat per workspace). The tiers are Starter (free), HQ Workforce, and Enterprise (custom pricing). It covers team features and cloud-run intelligence like meeting signals extraction. Humans are always free, with no per-seat fees. See pricing.

Knowledge base: The organized collection of written knowledge HQ draws on: markdown files under each company's knowledge/ folder and your personal space. It grows as you work. See Teaching HQ.

Knowledge garden: The practice of tending your knowledge base like a garden: pruning stale content, merging duplicates, fixing inaccuracies. Run /garden to audit a company's content.

Learn (/learn): The command that turns a finding into durable memory: a rule becomes a policy, an insight becomes a knowledge note, each filed in the right scope (personal, company, or repo). The engine behind "HQ gets smarter over time".

Meeting notes: HQ-native notes, transcripts, and recaps from recorded meetings, browsable with /meeting-notes. See Meetings and signals.

Ontology: A company's structured map of the people, products, projects, and concepts that matter to it, and how they relate. Read it with /ontology.

Outpost: A hosted HQ machine in the cloud that runs your scheduled jobs (set up with /schedule) and can host apps (with /outpost-host).

Personal overlay: Your private personal/ folder: profile, skills, policies, and projects that belong to you rather than to any company, and that survive HQ upgrades. See The HQ mental model.

Plan mode: A safety mode in Claude Code where the assistant proposes a step-by-step plan for your approval before changing anything. Different from the /prd command, which creates a project PRD.

Policy: A standing rule HQ follows automatically, stored as a small file scoped to a company, a repo, or you personally ("never email invoices without approval"). Created via /learn. See Teaching HQ.

PRD: Product Requirements Document, the plan file at the heart of a project, listing the goal and the individual user stories with their acceptance criteria. Created by /prd or /deep-plan. See From idea to shipped.

Project: A piece of work with a goal and a finish line, living under companies/{company}/projects/ or personal/projects/. See Companies, projects, and repos.

qmd / Search: HQ's search engine, which indexes your knowledge, policies, projects, and threads so you can find anything with /search. See Finding anything.

Ralph Loop: HQ's core working method: pick one task, plan it, execute it, verify it against acceptance criteria, commit, then move to the next task with fresh context. Named loops of small verified steps beat one giant unverified push.

Repos: The repos/ folder holding actual git code repositories, split into repos/public/ and repos/private/. Companies point to repos; code lives here.

Secrets: Credentials (API keys, tokens) stored encrypted and injected into commands only when needed via /hq-secrets, never pasted into chat. See Connecting your tools safely.

Session: One conversation with your AI assistant, with a limited working memory. Sessions are disposable by design; what matters gets saved to disk. See Sessions, checkpoints, and handoffs.

Share link: A secure, single-use link created by /hq-share that grants temporary access to a vault path (it expires after about 15 minutes). For permanent access, share directly with a teammate instead. See Sharing your work.

Signals: The useful items HQ extracts from meetings: action items, decisions, commitments, risks, and key points. Read yours with /signals. See Meetings and signals.

Skill: The instruction file that implements a command or a worker capability. You type commands; skills are the recipes behind them. See Commands vs skills vs workers.

Startwork (/startwork): The session opener: it surfaces your last handoff, active projects, and smart next options so every session starts oriented. See The daily loop.

Sync: The bidirectional exchange that keeps a cloud-backed company identical across your machines and your teammates'. The HQ app runs it automatically, or you can run it on demand with /hq-sync. See the sync guide.

User story: One small, well-defined task inside a PRD, with its own acceptance criteria ("US-001: add a contact form"). Small enough to finish and verify in one focused pass.

Vault: Your encrypted cloud file store, holding synced company content and shared files. See or change who has access to vault files with /hq-files, or fix a file you can't find or open with /hq-access. See HQ Cloud explained.

Worker: A persistent AI specialist you create with /newworker and run with /run {worker} {skill}. A named team member with its own skills, context, and guardrails that improves with use. See Commands vs skills vs workers.

Common questions#

If you get stuck#

If a definition here doesn't match what you're seeing in your HQ, your install may be older (run /update-hq) or the docs may be wrong, in which case /hq-bug reaches the team.