Glossary

Every word Player One uses that is not ordinary English, in one line each, with the page that explains it properly.

• Account. One connection to an AI provider, either a subscription you sign in to or an API key you paste. One provider can hold several. See Models and providers.
• Active agents. The page listing the conversations running this second, with a queue panel and a control that stops one without losing the chat. See Agents.
• Admin. The second menu, opened from the bottom of the sidebar, holding everything you configure or supervise about your installation. Back to Player One returns you to your chats. See Your installation.
• Agent. A named worker with one role, its own files, its own memory and a bounded set of tools. You add more as the work grows and call them whatever you like. See Agents.
• Agent Sharing. The Admin page where you share one of your agents with another Player One installation, use one shared with you, or open a sandboxed agent to anyone with the address. See Share an agent.
• Alerts. The Admin page where you say which channels each urgency level uses, test that they work, and read what Player One has already asked you. See Channels and communication.
• All projects. A browsing mode rather than a project: every list widens to the whole installation, and anything that must be filed somewhere cannot be created while it is on. See Organising with projects.
• API key. A credential from an AI provider that bills you per token rather than by the month. Keys you issue so your own software can call your installation are a different thing, described under LLM endpoint. See Models and providers.
• Approval. A point where work stops and waits for you before it goes on. Jobs hold a gated phase; a flow pauses where its description says a person decides; a command an agent's permission mode does not cover asks with Once, Always and No. See Stay in control.
• Audience. Whether a Memory Space is internal, meaning you and your agents, or public, meaning a customer or a bot answering one may read it. See Memory.
• Automations. The sidebar group holding Active, Flows, Jobs and Scheduled: everything your installation is doing without you sitting there. See Jobs.
• Backup. An export of chosen work as an archive you download, which can then be imported into another installation. See Back up and move your data.
• Beta. The label on something newer and less settled. Live voice calls carry a beta notice the first time you press the button. See Voice.
• Browser profile. A named, remembered identity in Player One's own browser: the cookies and sign-ins for one account on the sites it uses. One profile is one identity, a project can have a default, and a profile is never a password store. See The browser.
• Build. The compiled result of one version of a flow's description. A run uses the latest accepted build, and changing the description means building again. See Flows.
• Bundle. A project bundle: one whole project in a single download, with its agents, workspaces, Memory Spaces, flows, jobs, scheduled tasks and contact lists. See Back up and move your data.
• Cache. The provider holding the unchanged start of a conversation ready to reuse, and charging that part at a much lower rate. A cache is short-lived and rebuilds itself after a long gap. See Tokens, cost and caching.
• Channel. A way Player One reaches you and you reach Player One: Telegram, WhatsApp, Slack, Discord, your email, and the web app itself. The same assistant is behind every one. See Channels and communication.
• Chat. One continuous conversation with one agent, listed in the sidebar and reachable by its own link. See Chat.
• Cloud sync. A folder from a cloud drive kept in step with a folder inside a workspace, in one direction or both. See Workspaces and files.
• Compaction. Summarising the older part of a conversation so the chat keeps fitting, marked in the thread by a Context compacted divider. Nothing is deleted; what changes is what your assistant works from. See Conversations.
• Composer. The message box at the bottom of a chat, with the paperclip, the microphone, the model chip and the ring that shows how full the conversation is. See Chat.
• Contact. Someone Player One may need to reach, with a row for each way of reaching them and an order of preference among those rows. See Channels and communication.
• Contact list. A named address book. The same address can belong to a different person in each list, which keeps a guest list apart from your main directory. See Channels and communication.
• Context. Everything an agent can read at once: the whole conversation, the files it opened and its own replies. The limit is the context window, and it differs by model. See Conversations.
• Coordinator. The agent that plans a job, splits it into tasks and picks who does each. It plans and checks; it never does the work itself. See Jobs.
• Dashboard. The overview page: what is running, which agents have been busy, and anything wanting your attention. It is also where a dead provider sign-in raises its alert. See Your first hour.
• Devices. The Admin page listing every device paired with your installation, where you pair a new one or revoke one you no longer use. See Your installation.
• Early access. Something that works and is still being finished. Flows carry the label, so give a new one a few supervised runs before you rely on it. See Flows.
• Player One Provider. The transcription service included with Player One. It turns your voice messages into text from the start, with nothing to connect, and does nothing else. See Voice.
• Sidekick. The small app for your own computer, in the tray or the menu bar, that connects it to your installation so agents can see its screen, work in its programs, run commands and reach the folders you share, as far as you allow per computer: Off, Observe only, Change things or Full access. It also comes as a browser extension that lets agents work in the web pages you allow. See Sidekick.
• Effort. How hard a model thinks before it answers, on the models that allow the setting. It is separate from which model you are on, and can be set for one chat, on each model entry in a routing chain, or as a default. See Models and providers.
• Elevation. Permission for a command to run as the machine's administrator. It is granted to a person on a named channel rather than to an agent, and what a run inherits from it is availability: a command still runs unelevated unless it asks. See Stay in control.
• Engram. The curator that classifies and organises your Memory Spaces while you are not working: it files each note by meaning into a linked graph, keeps a four-tier recall index that follows use, and tends its own structure. Also the Admin page showing what it has queued, done and scheduled. See Engram.
• Failover. Moving to the next account, and then to the next model, when one cannot serve. The order of the cards on the Providers page is the order it tries. See Models and providers.
• Files. The sidebar row under Manage that opens your list of workspaces, and the first tab inside each one. The things it lists are workspaces. See Workspaces and files.
• Flow. One of your business processes, written down in plain English and compiled into software, so the hundredth run follows the same steps as the first. See Flows.
• Fold. The memory store behind search by meaning, built for proto-holographic recall: a fragment pulls in the decisions, files and history connected to it, weighted by what you use. Also the Admin page showing every index, its health and what it has cost. See Fold.
• Folder. A named group of chats in the sidebar, inside one project, and nestable. See Conversations.
• Forge. Where new agents come from: a catalogue of ready-made specialists that opens beside the chat, and the one-pass build your main assistant runs once you say go. See The Forge.
• Grant. Permission for one agent or one project to use one vault item, with a scope and a lifetime, revocable whenever you like. See Vault and sandboxing.
• Help. The panel that slides in beside almost every page, opened by the round button in the corner or by Explain more beside a page title. It is your own assistant, not a support desk, and it knows which page you are on. See Help inside Player One.
• Helper. A background agent given one self-contained slice of work while you carry on in the chat. Its answer comes back as a card, and it cannot see your conversation, so its brief has to stand on its own. See Agents, delegation and team chat.
• Hide. Removing a chat's row from the sidebar without deleting it. The chat still opens and search still finds it. See Conversations.
• Import. Bringing an archive exported from another installation into this one, picking what to take and reviewing the plan before it is applied. See Back up and move your data.
• Installation. Your own running copy of Player One at your own web address, holding your chats, agents, files, memory and connected accounts. Nothing in it is mixed with another customer's. See Your installation.
• Instructions. The standing rules an agent reads on every turn, at three levels: site-wide, then the project's, then the agent's own, with the narrowest having the last word. See Organising with projects.
• Job. A piece of work too big for one conversation: your brief, a plan, a team of agents, phases that run in order, and a record of everything that happened. See Jobs.
• Knowledge. What a project may search, connected on its Knowledge tab, and separately the tab on a workspace that says whether its files can be found by meaning. See Organising with projects.
• LLM endpoint. The address your installation publishes so your own software can call the AI models you have connected, with keys you issue, watch and revoke. See Use Player One from your own apps.
• Main assistant. The front door: the agent that knows you, answers most things itself, hands work to whoever is better placed, and is the only one that can create or change another agent. See Agents.
• MCP. The Admin page where you connect a coding agent on your own computer, such as Claude Code, Codex or Cursor, to your installation with a key of its own. See Build with it.
• Memory. What your agents have written down and can find again: facts, preferences and decisions, kept outside any one conversation. See Memory, explained.
• Memory Space. A named collection of notes with a stated purpose, an audience and its own guidance on what belongs in it. Every agent has one, and you create shared ones. See Memory.
• Messages. The feed of agents talking to each other across your jobs, including anything addressed to you. It has no menu row: you reach it from the notification bell or the Dashboard. See Agents, delegation and team chat.
• Messaging. The Admin page holding one card per messaging and email connection, with its live state and the buttons to set one up, repair it or disconnect it. See Channels and communication.
• Model. The AI brain that actually answers. Each connected account brings a set of them, and you can overrule the choice for one chat, one agent or the whole installation. See Models and providers.
• Needs input. The state a job enters when it has asked you a question and stopped until you answer. See Jobs.
• Notification bell. The list inside Player One of everything it has asked or told you. It is included in every urgency level, so there is always one place with the whole record. See Channels and communication.
• Permission mode. What one agent may run on your installation without asking you: Conservative, Most things or Everything, set on the agent's Advanced > Permissions tab, with a site default for agents without their own. See Stay in control.
• Phase. A group of a job's tasks that runs together. Phases go in order, with work happening in parallel inside each one, and a gated phase waits for your approval. See Jobs.
• Phone line. A real telephone number given to one agent through a telephony provider, so it can take and make calls. A line is registered once by asking your assistant; each call is transcribed into a chat. See Voice.
• Plan. The coordinator's living document for a job, rewritten as the work goes and shown on the job's Plan tab. See Jobs.
• Preview pane. The panel to the right of the conversation where a document, a page, a flow or a browser window opens and refreshes as it changes. A Preview button in the chat header opens it with your files, and your agent opens it to show you something, so you ask rather than click. See The preview pane.
• Profile (browser). A named browser identity that keeps cookies and site state, so a login done once keeps working. One profile is one identity, and two accounts on the same site need two. See Log in for it once, let it work from then on.
• Progress log. The panel behind the line at the foot of a running reply, What's happening while it runs and What happened once it has finished, holding every step with what it was given and what it returned. See Chat.
• Project. The boundary round one initiative's work, grouping its chats, agents, workspaces, Memory Spaces, jobs, flows, scheduled tasks, contact lists and skills. See Projects.
• Provider. A company whose AI models Player One can run on. Several can be connected at once, and the order you drag them into decides who answers first. See Models and providers.
• Provider (trigger source). A connected outside source that watches for events and hands them to a flow, listed on the Flows page's Providers tab. See Set up a trigger.
• Reference (vault). The handle an agent holds instead of a secret. It names the item and carries neither the value nor permission to use it. See Vault and sandboxing.
• Release notes. What changed in a new version, which your assistant walks you through when your installation updates rather than leaving you to notice. See Your installation.
• Request (vault). An agent's ask for access to a vault item it has not been given. It waits on its own tab and arrives as a card in the chat that raised it. See Vault and sandboxing.
• Resources. The Admin page showing the health of the machine your installation runs on: memory, processor, disks and thirty days of history. It watches and changes nothing. See Your installation.
• Run. One firing of a scheduled task, recorded with how it went. Open the task for its run history. See Scheduled tasks and watchers.
• Run (flow). One execution of a flow, with its own inputs, its event log and its result, drawn as a map with the current step lit up. See Flows.
• Runtime. The Admin page carrying your installation's connection status and version numbers, the Advanced and Development switches, and the buttons that clear caches or restart. See Your installation.
• Sandboxed agent. An agent you can put in front of customers. It knows only what you gave it, acts only through the narrow tools you built, has no vault and no general search, and never joins a team or a job. See Vault and sandboxing.
• Scheduled task. One agent running one instruction on a clock, delivered somewhere you will see it. See Scheduled tasks and watchers.
• Search. The page behind the box in the top bar, which finds chats, agents, jobs and files together and can be narrowed to one kind. See It remembers.
• Search by meaning. Finding something by what it is about rather than by the words in it, so a question phrased nothing like the note still turns it up. See Memory, explained.
• Secret. Anything that must not be seen: a password, an API key, a private key, a certificate or a secure note. Secrets live in the vault and are never displayed. See Vault and sandboxing.
• Secure browser window. A real browser opened beside the chat so you can sign in to a site yourself. The password never passes through the agent. See Log in for it once, let it work from then on.
• Skill. A short written playbook for doing one kind of thing well, kept outside any conversation so any of your agents can pick it up. See Skills.
• Skill notes. The corrections that build up against a skill and attach every time it loads, winning over the original wording. Open a skill on Admin > Skills and its notes are listed under the text, grouped by who collected them. See Teaching it.
• Sleep. The background pass in which Engram organises a Memory Space: filing what has arrived, merging it with what was there and keeping the Space's own guidance current. See Memory, explained.
• Speak back. The setting that has replies read aloud as well as written. It pauses by itself during a live call. See Voice.
• Specialist. An agent whose role points at one kind of work, so its memory of that work compounds and it gets better at the job than a generalist would. See Agents, delegation and team chat.
• Strengths. The short description of what an agent is good at. It is what a coordinator reads when it picks someone for a task, so plain words pay off. See Agents.
• Suspended. A flow run that is waiting for a person or a timer: paused rather than finished, and it picks up from the step it stopped at. See Flows.
• Take control and Hand back. The two buttons that move the keyboard between you and the agent in a secure browser window. See Log in for it once, let it work from then on.
• Task. One unit of work inside a job, assigned to one agent and shown on the board as Pending, In Progress or Finished. See Jobs.
• Team. How your internal agents are organised: who leads, who does what, and who reports to whom, written as a team page every internal agent reads. Set up and changed by telling your assistant; sandboxed agents stay outside it. See Agents.
• Team chat. Agents talking to each other, shown in your conversation as cards attributed to the agent that sent them rather than to you. See Agents, delegation and team chat.
• Token. The unit AI providers count and charge in: a chunk of text, roughly three quarters of a word. Both what you send and what comes back are priced this way. See Tokens, cost and caching.
• Trash. Where a deleted chat or Memory Space waits, so you can preview it, restore it, or remove it for good. See Conversations.
• Trigger. What sets a flow going: a schedule, a message arriving on one of your channels, an event from a connected source, or you pressing run. See Set up a trigger.
• Update. A new platform version. A red Required update banner appears and your assistant opens a conversation that goes through each release and repairs anything the new version needs. See Your installation.
• Urgency. The level on a question or an alert, Urgent, Normal or Quiet, which decides the channels it goes to. See Channels and communication.
• Usage. The Admin page answering what this is costing you: tokens and money over a window you pick, broken down by model and by agent. See Your installation.
• Use as knowledge, Use and update. The two levels at which a Memory Space or a workspace is connected to a project: read and search only, or write to it as well. See Organising with projects.
• Vault. The encrypted store in your installation for keys, passwords, private keys, certificates and secure notes. Stored values are never displayed, to an agent or to you. See Vault and sandboxing.
• Voice call. A live spoken conversation with an agent, started from the phone button in the composer. You can talk over a reply and it stops and listens. See Voice.
• Voice message. A recording made with the microphone in the composer, transcribed on arrival, so the transcript is your message. Audio you attach with the paperclip is left alone. See Voice.
• Voice orb. The circle on screen during a live call, pulsing with the audio and captioning the words as they are spoken. See Voice.
• Watcher. Something armed to notice a change and tell you once, rather than run on a clock: a file that moves, a value that shifts, an outside system that posts in. See Scheduled tasks and watchers.
• Web search. How your agents search the web: through a search service chosen on the Web Search tab of Agent Models, or automatically. See Models and providers.
• Webhook. An address on your installation that outside software can post to, which either turns the message straight into a message to you or hands it to an agent for one turn. See Scheduled tasks and watchers.
• Wide view. The preview pane at about three quarters of the screen, with the sidebar tucked away and the chat narrowed to a column beside it. See The preview pane.
• Worker. An agent that takes routine work off the main conversation, and the kind of agent a coordinator hands a task to inside a job. See Agents.
• Workspace. A named folder of files you and your agents share, with its own cloud folders, its knowledge setting and its project connections. The sidebar row that lists them is Files. See Workspaces and files.

Anything here that turns out to be a whole subject rather than a word has a page of its own: Meet the parts describes each part of Player One in turn, and Working with Player One explains the ideas behind them.