​

General Settings

The assistant editor’s left menu has six tabs: General, Provider, Instructions, Knowledge, Integrations, Context. This page covers General — your assistant’s identity and how it starts.

​

Name and Description

Name is the assistant’s display name, shown throughout the platform. Description is a short summary of what it does. Both feed the assistant’s own AI: the clearer the name and description, the better it can improve its own prompt when AI Prompt (Context tab) is enabled. Guidelines:

• Be specific about the assistant’s function — “Meeting Notes Assistant” beats “Assistant 1”
• Keep the name concise (2–5 words)
• Use the description for anything a teammate would need to know before reusing this assistant

---

​

Timeout (seconds)

Maximum time allowed for the assistant to respond, in seconds. A value of 0 applies the default timeout of 300 seconds. Raise it for assistants that do longer multi-step work.

---

​

Response Mode

A single toggle, not two named modes. Its two states:

• Wait For Response (default) — the workflow pauses until this assistant completes and returns a result. Use this when a later step depends on the assistant’s output, or the caller is waiting synchronously (chat, an API call, a form submission).
• Continue Immediately — the workflow continues without waiting; the assistant runs in the background. Use this for notifications, logging, or any task that doesn’t feed back into the flow.

---

​

Prompt

The Prompt field on this tab is empty by default — it does not ship with a placeholder value.

• If the assistant is directly connected to a trigger, the triggering message is passed in automatically; you don’t need to fill this in.
• If the assistant sits behind another step, map the value in explicitly from that step’s output.

---

​

Global Instructions and Global Knowledge

Two independent toggles, both off by default:

• Global Instructions — includes your account-level shared instructions (configured in Account Settings) in this assistant’s behaviour.
• Global Knowledge — includes your account-level shared knowledge sources in this assistant’s context. This is a single all-or-nothing switch: there is no per-source picker at the assistant level. See Knowledge for what account-level knowledge contains and who can see it.

---

​

Escalations

Choose which account users are notified when this assistant needs human escalation.

---

​

Instructions live on their own tab

The assistant’s core behaviour is not configured on General — it’s a separate top-level Instructions tab, with three sub-tabs:

• Assistant Instructions — the main field defining role, personality, and behaviour. Use the Generate Instructions (or Modify Instructions) button to draft it with AI assistance, then edit directly.
• Acceptance Criteria — an optional list of guidance strings for response validation and refinement. Empty by default.
  Acceptance Criteria are not currently enforced — nothing in the response pipeline reads them yet. Anything you put here is documentation for your team, not a runtime check.
• Context Variables — an optional free-form key/value map injected into the assistant’s prompt (for things like a username or an integration-specific parameter). Empty by default.
  Context Variable values are not sanitised and are injected verbatim — they can enable prompt injection. Use only with trusted input, and never store credentials or secrets here.

---

​

Next Steps