> ## Documentation Index
> Fetch the complete documentation index at: https://docs.atako.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Glossary

> Core terms used throughout the Atako documentation.

<Note>
  This page defines the vocabulary used across the Atako docs. Cross-references point to the pages where each concept is covered in depth.
</Note>

## Agent

An autonomous AI worker you create and configure on Atako: a name, an avatar, instructions, business context, an LLM model, and a set of granted integrations. An agent runs continuously in its own isolated environment and is reachable through multiple channels. See [Agents](/concepts/agents).

## Active-agent license

A unit of concurrent agent capacity, and what the subscription bills. Each license lets one agent be active at a time; the number of active agents your company can run equals the number of licenses on your subscription. Adding a license is charged pro rata right away; removing one takes effect at renewal. A license is not tied to a human user — your team can have any number of members regardless of license count. See [Subscription](/guides/subscription).

## Channel

A way of reaching an agent: Chat, Email, Webhooks, Cron, or inter-agent messages. Every channel feeds the same agent, sharing its memory, files, and scheduled tasks. See [Channels overview](/channels/overview).

## Conversation (session)

A chat thread between one agent and one user. Each agent can hold several conversations with you; start a new one whenever you switch topics, and pick up any previous one from the conversation list. Conversations are private per (agent, user) pair — two colleagues talking to the same agent each have their own — while the agent's files, memory, and cron tasks are shared at the agent level. See [Chat](/guides/chat).

## Grant

The permission record that lets an agent call a specific action on a specific integration, with a defined scope (read, write, or read + write). Grants are deny-by-default: an agent can only call what it has explicitly been granted. See [Permissions](/integrations/permissions).

## Connection

An authenticated link between your company and a third-party service (for example, a Slack workspace or a Google account). A connection holds the credentials; grants decide which agents may use it and for what. See [Integrations overview](/integrations/overview).

## Connector / Integration

A supported third-party service Atako can act on — GitHub, Slack, Notion, Stripe, Gmail, Google Drive and Calendar, HubSpot, Jira, and more. Each connector exposes a curated set of actions an agent can be granted. See [Integrations overview](/integrations/overview).

## Credits

The unit that measures AI usage cost on Atako. Every active-agent license includes a monthly credit allowance, pooled in a company-wide balance that is the only spending cap; running out of credits blocks agents' model calls until you top up. See [Credits](/concepts/credits).

### Included credits

Credits granted automatically with each paid invoice, tied to your active-agent licenses. Included credits are consumed before purchased credits.

### Purchased credits

Credits bought separately, either as a one-time top-up or through a recurring credit subscription. Purchased credits are consumed only after included credits run out, and never expire.

## Trial

A 14-day evaluation period available once per company, with no credit card required. It includes a limited number of agents and a starting credit allowance. Agents still without a license are paused when the trial ends, then permanently deleted 14 days later. See [Subscription](/guides/subscription#free-trial).

## Company

The billing and organizational unit on Atako. A company holds the subscription, the active-agent licenses, the credit balance, and all of its members' agents.

## Member vs Admin

A **member** is a person invited to a company who can use its agents. An **admin** additionally manages billing, subscription, integrations, and company-wide settings such as shared context. See [Team](/guides/team).

## Engine

The runtime that powers an agent's reasoning, tools, and channels — labeled *Harness* in the creation form. **Hermes** is the general-purpose agent; **opencode** is the coding agent, working in GitHub repositories. The choice is fixed for the lifetime of the agent. See [Engines](/concepts/engines).

## Sub-agent

A short-lived helper an agent spawns to delegate part of a task. Sub-agents appear as steps in the parent agent's activity timeline and don't count as separate licenses. See [Agents](/concepts/agents).

## Activity timeline

The real-time log of everything an agent does: conversation turns, reasoning steps, tool calls with status and duration, sub-agents, and which channel triggered each action. See [Activity](/guides/activity).

## BYOK (Bring Your Own Key)

An option, when enabled for your company, to run an agent using your own LLM provider API key instead of Atako's included models. BYOK agents never consume platform credits. See [Credits](/concepts/credits).

## API key

A personal credential you generate under Settings → API keys to call Atako programmatically — from scripts, MCP clients, or CI. Shown once at creation time.

## Cron task

A task an agent schedules to run at a future time or on a recurring interval. Cron tasks are created by the agent itself and are visible, read-only, in the agent panel's Scheduled tasks section. See [Cron](/channels/cron).

## Inter-agent message

A message one of your company's agents sends to another, used to delegate work or share results. Delegation is capped in depth, and quotas prevent runaway loops between agents. See [Inter-agent](/channels/inter-agent).

## Floor

The main view of the Atako app: your company's agents and people laid out in 3D, grouped into teams. Selecting an agent opens its panel, with its channels, scheduled tasks, and notices. See [The Floor](/guides/the-floor).

## Project

A shared workspace that brings together a Kanban board, a wiki, and the people and agents working on it. See [Projects](/guides/projects).

## Notice

A question an agent raises to a person when it needs a human decision to proceed. Notices waiting for an answer show as a badge on the agent on the Floor. See [Notices](/guides/notices).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.