Glossary¶
One line per term. The link goes to the page where the term is defined in full.
| Term | Meaning |
|---|---|
| Agent | A representative of your business that talks to people on messaging channels. Made of Behavior, Knowledge, Capabilities, and Channels. |
| Agent template | A complete starting point for a whole agent, curated by Anychat, carrying everything except identity. |
| API key | A workspace-scoped credential for the Agent API, carrying the scopes chosen when it was minted. Shown once, at creation. |
| Automated Outreach | A named rule that sends an Outreach to every new qualifying row on one linked Google Sheet tab. An agent keeps one per lead source. |
| Behavior | The document you write to specify how an agent acts: Identity, Personality, Guardrails, Rules, Goals, Playbooks, and message templates. Called the Definition before September 18, 2026, which is still the name of the /v1 resource. |
| Binding | The connection between one agent and one channel. Managed bindings are provisioned and operated by Anychat; unmanaged bindings use a provider account you run. |
| Business info | Typed facts about the business itself: address, hours, amenities, price range. |
| Campaign | One Outreach sent to a list of People you pick, staged and started as a unit. |
| Capability | Something the agent can do beyond talking, which you switch on and configure: Scheduling, People, Notifications, custom integrations. Knowledge is not one. |
| Case unit | The measured weight of one Test result, used for quotas and metering. |
| Channel | Where the agent speaks: RCS, Apple Messages for Business, WhatsApp, SMS, and others. |
| Channel identity | One person's address on one channel, such as their number on RCS, with an opaque handle called its ASID. A Person is found only through the identities it holds. |
| Channel line | The one-line description of the current channel the agent reads each turn. |
| Collection | A named group of Things that controls how their cards look and can be sent as one carousel. |
| Concept | A self-playing demo conversation generated from a prospect's website, published as a shareable link. |
| Conversation | Every message exchanged between one person and one agent, over all time. |
| Custom integration | A capability backed by your own HTTP endpoint, described in your words. |
| Default Playbook | The Playbook that is active when no other fits, and where the agent returns when a task is done. |
| Early access | The list of phone numbers that may reach an agent on a channel before that channel is launched. |
| Emulator | The browser page that renders a conversation the way a handset would. In the console under Test, and as a standalone link for someone without a console account. Always Sandbox. |
| Fact | A question and the answer you have confirmed, read as authoritative on every turn. |
| Flow | A fixed path carried by a Playbook: steps the platform walks in order, with the agent filling in each step's blanks. |
| Goal | The one outcome a conversation works toward, assigned by Playbook, by Outreach, or as the agent's default. |
| Guardrail | A limit in the Behavior document the agent reads and follows with judgment, as opposed to a Rule, which the platform enforces. |
| Knowledge | What the agent knows: Sources, Facts, Things, Media, and business info. Present on every agent and never a Playbook's Need. |
| Lint | The check that runs on every save and reports inconsistencies in the Behavior document. |
| Live | A real person on a real device over a real channel, with real effects. The other world is Sandbox. |
| Media | The pictures and videos the agent may show, referenced by id rather than by URL. |
| Message template | Exact copy you have reviewed, sent verbatim with placeholders filled in. |
| Outreach | The reusable reason an agent reaches out: an opener, a Playbook, a Goal, and a class. It names no audience. |
| Person | The record an agent keeps of one human it talks to. |
| Persona | A Person record that lives in the sandbox, with a starting point. A persona is who talks to the agent in Sandbox, and can be the recipient of anything outbound. |
| Playbook | What the agent does in one kind of situation: name, when-to-use, body, and the capabilities it uses. |
| Readiness | Whether a capability, and so each Playbook that uses it, is set up and available. One answer shared by the console and the runtime. |
| Representative | A person on your team the agent books calls with. |
| Review | A model's reading of the whole agent, on request, reporting the problems a careful colleague would. |
| Role | A member's level in a workspace: viewer, editor, admin, or owner. Each includes the ones below it. |
| Rule | A limit the platform enforces before or after the agent's turn. The agent never sees it. Checked Rules are deterministic and absolute; Judged Rules are read by a fast model before the reply is sent. |
| Sandbox | The world in which a persona talks to the agent through the emulator, a Test, or a script: reads are real, anything stored is kept in the sandbox partition, anything that would leave Anychat is captured and shown. Never counted. |
| Scope | A named permission on an API key, such as agents:read. A key never exceeds the role of the member who minted it. |
| Send | One Outreach message to one person. The row you inspect, count, and cancel. |
| Service window | The period after an inbound message during which a channel lets the agent reply freely. 24 hours on RCS and WhatsApp. |
| Session | One sitting within a conversation: messages with less than 30 minutes of idle time between them. The unit of reporting. |
| Source | Material the agent answers from: an uploaded file, an imported page, or pasted text. |
| Starting point | What a persona's record and conversation are when it is made, and what Reset returns them to. preset in the API. |
| Standing instructions | Short house rules in Identity that apply in every conversation, whatever the Playbook. |
| Test | A conversation with expectations attached, run in Sandbox without a human, on demand or on a schedule. |
| Thing | One record in your catalog: a product, a room, a rental, a service. Presented as its own card. |
| Turn | One inbound message and the agent's reply to it. |
| Turn record | What Anychat keeps for every turn: what fired, what was read, which Playbook was chosen, what the agent did and said, and what was sent. |
| Uses | The list of capabilities a Playbook acts through, shown as Needs in the console. |
| Workspace | The account that owns your agents, People, channels, keys, and branding. Called an org in the Agent API. Nothing is shared between workspaces. |