Try Esan

SoonAssistants and agent-to-agent are coming soon. This page describes how they will work.

Assistants

Esan isn't one assistant — it's a roster. An assistant is a named agent you shape for a specific role: a researcher, an editor, a sales assistant, an analyst. Each one has its own role and knowledge, and all of them draw on what Esan remembers about you.

Why use assistants

  • Sharper results. An agent told it's "a B2B SaaS researcher" frames work the way that role would.
  • One memory about you. Every assistant reads the same memory, so what you tell one, the others know too.
  • Reusable. Set one up once and delegate to it again and again.

Create an assistant

  1. Go to Assistants and create a new one.
  2. Give it a name and a short role ("a financial analyst for early-stage startups").
  3. Optionally add a persona — tone, focus areas, and things it should avoid — so its output matches your expectations.

Configure an assistant

Open any assistant to tune it:

  • Persona — name, role, tone, focus areas, and limits.
  • Knowledge — a long-term document the assistant always has on hand (style guides, background, rules of thumb).
  • Connectors — which of your connected apps this assistant may use.
  • Skills — which reusable playbooks it can run.
  • Files — documents every session with this assistant can read.
  • Voice — the voice and style used when reading its replies aloud.

Talking to an assistant

Open an assistant and start a chat, or mention it from the composer to hand a task to the right teammate. Bring another assistant into a conversation for a second opinion when you want one.

Tips

  • Keep each assistant focused. A few sharp roles beat one do-everything agent.
  • Put durable context in Knowledge, not in every message.
  • Use per-assistant connectors to keep access tight — give your researcher web access without handing it your email.

Next: Memory & personalization →

Was this page helpful?