Core Concepts
Workprentice has a small, consistent vocabulary. Once you know these terms, every screen in the product reads the same way. Each entry below gives the definition and an example of when you’ll reach for it, with a link to the full page.
The actors
Section titled “The actors”Organization
Section titled “Organization”Your company. An organization holds members, workspaces, shared skills and integrations, usage reporting, and the audit log. Creating one is optional — you can use Workprentice solo and add your team later. Once you have one, org admins can set organization-wide defaults (a skill every workspace inherits, a guardrail no workspace can loosen) and review usage and activity across the whole company. See Organizations and workspaces.
Workspace
Section titled “Workspace”A team-sized scope inside an organization. Agents, assignments, sessions, and work surfaces all live in a workspace, and members of the workspace share them. A sales team might keep its pipeline agents, deal trackers, and weekly reports in a “Sales” workspace while the support team runs its own — same organization, separate working contexts.
An AI identity with a personality, skills, tools, and guardrails. Your personal agent — the “me” agent pinned at the top of the sidebar — is created for you automatically and works across your workspaces. You can create additional agents for specific jobs — a support triager, a pipeline analyst, a docs writer — and keep them private or share them with the workspace, choosing what teammates may do with each (view, use, or edit). See Creating and configuring agents.
Personality
Section titled “Personality”The character and standing instructions an agent carries: how it writes, what it cares about, how it approaches work. A pipeline analyst’s personality might read “concise and numbers-first; flags risks early; never pads a report.” You can edit it directly or let the agent refine it as it learns.
The work
Section titled “The work”Assignment
Section titled “Assignment”A job described once and run many times: instructions, parameters, and the tools it needs. Assignments always run on demand, and can also run on a schedule or from a webhook — a weekly pipeline report is a scheduled assignment; “enrich each new support ticket” is a webhook-triggered one. If a conversation turns out to be a recurring need, Make recurring distills it into a draft assignment. See Assignments and Schedules, webhooks, and triggers.
Session
Section titled “Session”One conversation or one run of an assignment. Sessions record every thought, tool call, and result as the agent works — the full audit trail, streaming live and kept permanently. You can stop a running session or send a message mid-run to steer it, and sessions can be private to you or shared with the workspace. When a Monday report looks off, the session is where you find out why: open the run and read exactly what the agent did. See Sessions.
Your triage queue. When an agent needs a decision, hits a blocker, or has something worth your attention, it lands here with a priority — Urgent, Important, or Normal — instead of failing silently. A typical item: an overnight run found its credentials expired and asks you to reconnect. You reply in the thread, forward the item to another agent, or snooze it for later. See Inbox.
Work Product
Section titled “Work Product”A durable deliverable produced in a session: documents, spreadsheets, PDFs, charts, and images. Work products live beyond the conversation that produced them and are collected on the Work Products page, where each one links back to the agent and session that made it — so the reasoning behind any deliverable is one click away. See Work products.
The surfaces
Section titled “The surfaces”Data Table
Section titled “Data Table”A structured table with typed columns — text, numbers, dates, selects, and more — that agents can read and maintain alongside you. Where a recurring report produces a fresh document every run, a data table stays one current, queryable place: a competitor tracker an agent refreshes every Friday, or a rolling log of weekly metrics. See Data tables.
A kanban board of tasks the team and its agents share. Assign a task to an agent and its card shows the run status as the agent works. Boards can also adopt assignments, so recurring agent work shows up as cards the team can track alongside everything else. See Boards.
Knowledge Base
Section titled “Knowledge Base”Curated reference material — folders and files agents consult while working: a product handbook, a support playbook, a style guide. Every file keeps revision history, and where editing is allowed, agents help keep the content current. See Knowledge.
An interactive tool built in your workspace, versioned and publishable, that outlives any single session. Ask an agent to build a quote calculator or a team dashboard, review the preview, and publish it for the workspace to use. See Apps.
The abilities
Section titled “The abilities”Integration
Section titled “Integration”A connection to an external service — Slack, Salesforce, GitHub, Google Workspace, and dozens more — made through the open Model Context Protocol. Connections are credentialed at a scope: private to an agent, shared with the workspace, or shared across the organization. Once connected, the service’s tools simply appear in the agents’ toolkit; there is nothing to wire per assignment. See Integrations and the catalog.
Reusable know-how: workflows and expertise your agents apply on their own — a code review checklist, a report format, the way your team qualifies deals. Skills can be private to an agent, shared in a workspace, or published across the organization, so a practice written down once travels to every agent that needs it. See Skills.
Memory
Section titled “Memory”What an agent saves across conversations. Each agent keeps its own memory, reviewable and editable, and consolidates what it has learned while idle. Correct an agent in conversation — “actually, our fiscal year starts in February” — and the correction becomes memory that shapes every future run. See Memory and storage.
Guardrails
Section titled “Guardrails”The policies that bound an agent’s autonomy: network access, self-improvement, collaboration with other agents, and integration access. Set them per agent, and layer defaults at the workspace and organization level — restrictions flow downward, so a child level can tighten but never loosen what a parent set. See Guardrails.
How it fits together
Section titled “How it fits together”An organization contains workspaces. A workspace contains agents, which take on assignments. Every run is a session, which produces work products and may update data tables, boards, knowledge bases, or apps. Integrations and skills give agents their reach; memory makes them better over time; guardrails and the inbox keep you in control.
Traced through one example: your Sales workspace has a Pipeline Analyst agent with a weekly report assignment. Monday’s run is a session — you can open it and read every step — and it delivers a document as a work product while updating the deal-tracker data table. The agent reached the CRM through an integration, formatted the report the way a skill prescribes, and remembered last week’s feedback through memory. When the CRM credential expired, the run escalated to your inbox instead of guessing.
For the complete site index, see llms.txt.