Glossary

The words we use across this help center, in one place.

Agent. The Campfire bot in your Discord. It answers players’ questions about your game. You give it a name and face on the persona page.

Allowlist. The set of channels the agent is allowed to read and post in. Nothing is on until you add it. See channels it watches.

Answer mode. How a channel lets players reach the agent: by mention, by /ask, or automatically. Set per channel on channels and answer modes.

Auto mode. An answer mode where the agent replies to game questions on its own, without being mentioned or commanded.

Confidence. How sure the agent is of an answer. Low-confidence answers are surfaced in your analytics and feed the unmet-questions view.

Connected source. Source code or design docs you link so the agent can ground answers in them: GitHub, Notion, Google Docs, or Atlassian. See how the agent is trained.

Dashboard. The web app at dashboard.atcampfire.xyz where you configure the agent and read analytics. See dashboard.

Escalation. Routing a question the agent won’t answer, like account or billing, to a channel or role on your team. See escalation.

Grounded answer. An answer built on your game’s own materials rather than a guess. The core of what the agent does. See grounded answers.

Ingestion. Training the agent on your game so it knows its data: the published build, plus any source code and design docs you connect. Re-run after a patch. See how the agent is trained.

Persona. The agent’s name, avatar, and voice. See persona.

Spoiler line. The point of progression past which the agent blurs content behind a spoiler tag. See spoiler rules.

Unmet question. A question the agent couldn’t answer with confidence. The most useful signal in your dashboard. See unmet questions.