Glossary
Short definitions of every term used in the app and the docs, from action cell to workspace.
Updated 2026-10-13
Your company and plan
| Term | Meaning |
|---|---|
| Brief | What you tell Pythe about your company at the start, or the text you give an employee for a task. |
| Dossier | The confirmed facts about your company from onboarding. Every agent reads it. |
| Graduation | Moving from one stage to the next, once enough of your milestones are done. |
| Milestone | One piece of work in the plan. Also called a task. |
| Operating plan | The plan: stages, tracks, milestones and graduation criteria, stored with the workspace. |
| Stage | Where the company is on the seven-step ladder. |
| Track | A kind of work in the plan: Product, Growth, Revenue, Team, Fundraising, Operations or Customer. |
| Workspace | One company and one plan, with its own team, Library, connectors and limits. |
Agents and their work
| Term | Meaning |
|---|---|
| Computer | A cloud sandbox that belongs to one agent. See How agent computers work. |
| Definition of done | The “Done when” list on a tracked task. The work is checked against it. |
| Delivery verdict | The server’s judgement of whether a run delivered what was asked: delivered, partial, answered, no deliverable, delegated or failed. |
| Employee | An AI teammate hired from a role template. |
| Focus | A tool family (code, web, documents or research) added for one message only. |
| Hand-off | Giving work to Pytheus, an employee or yourself, as a tracked task; or one agent giving part of a job to another. |
| HQ | The Pytheus workspace at /cofounder: Ask, Work, Team and Plan. |
| Memory | The shared facts, preferences, decisions and findings the whole team reads before it starts. |
| Persona | How an employee works: their style and standards. |
| Pytheus | Your AI cofounder. On every plan, leads the team and has its own computer. |
| Role | A template an employee is hired from: Researcher, Outreach, Writer, Engineer, Analyst or Operator. |
| Run | One piece of agent work with a status, steps, a report and files. |
| Segment (part) | A further stretch of a long job, with its own turns, started from a compact summary of the earlier part. |
| Tracked task | A run with a brief, a definition of done, a channel, a progress log and a check. |
| Turn | One model call plus its tool calls. |
| Watch | The panel that shows what an agent is doing: Activity, Screen and Files. |
Work you can see and keep
| Term | Meaning |
|---|---|
| Action cell | A list item on a mission canvas that can run in place, with a button such as “Do this” and streaming output. |
| Canvas | The infinite board for thinking with your agents. Also the output page of a playbook mission. |
| Library | Everything saved: research, documents, notes, canvases and finished files. |
| Mission | A plan task (or a typed prompt) opened as its own workspace. |
| Needs you | The single list of everything waiting on you. |
Safety and money
| Term | Meaning |
|---|---|
| Approval | A card showing the exact action an agent wants to take outside Pythe. You approve, edit or reject. See Approvals. |
| Autopilot | A Premium setting, per employee and app, that skips the wait for a short list of actions. Never covers publishing a website. |
| Computer time | Minutes agent computers run, counted per month against your plan. |
| Credits | The allowance for AI work, counted from model use: 10,000 tokens is one credit. |
| Lease | A short claim a worker holds on a run or a plan build. If it lapses, another worker takes over. |
| Sweeper | The scheduled job that wakes stalled runs and closes dead approvals without anyone having the app open. |
Connecting your own tools
| Term | Meaning |
|---|---|
| Connector | A connected app such as Google, Slack, Notion, HubSpot, GitHub, GitLab, Stripe or PostHog. |
| Idempotency key | A value you send when retrying so the work starts once. |
| MCP | Model Context Protocol: how your own AI tools talk to Pythe. See The Pythe MCP server. |
| Scope | A permission on an API key: read context, author content, run actions or manage tasks. |