Docs

Agents

Beyond the default assistant, you can create specialized agents — each with its own identity, model, skills, and memory. Manage them in Settings → Agents, and summon one in the launcher with @agent-id.

Why create an agent

A purpose-built agent gives you consistent behavior for a recurring role — a code reviewer, a research analyst, a daily-standup writer. It always uses the right model, knows the relevant skills, and keeps its own memory separate from your general assistant.

What you configure

Each agent has several tabs:

Prompt

  • Display name and agent ID — the ID is the summon handle (@code-reviewer) and is fixed once created.
  • Description — what the agent is for.
  • Default model (optional) — overrides the conversation's model when this agent runs.
  • Soul — identity and personality, prepended to the system prompt.
  • System — operational instructions, applied after the soul.
  • Photo (optional) — an avatar (PNG/JPEG/GIF/WebP/AVIF, up to 10 MB).

Memory

A dedicated memory file the agent reads for context, editable here and visible on the Memory page.

Channels

Subscribe the agent to shared memory channels (e.g. #project-notes, #team-guidelines) so it draws on common context.

Skills

Pin a subset of your installed skills so this agent prioritizes them when matching prompts.

Schedules

List and manage the scheduled tasks bound to this agent — run now, edit, or delete. When a bound schedule fires, the task runs as this agent.

Summoning an agent

Type @agent-id in the launcher to start a conversation with that agent (for example, @research summarize this paper). Tasks bound to the agent automatically run under it.

Built-in agents

Agentastic ships with system-managed agents like Deep Research and Slide Creator. These power the corresponding apps and can't be deleted, but you can use them like any other agent.

Related