Editable prompts
Read what the assistant is told.
The instructions behind the chat, the composer and thread summaries sit in your workspace settings, word for word. Rewrite one and the next request follows it.
In short
What is a system prompt?
A system prompt is the standing instruction a language model reads before your request: the role to play and the rules to keep. Change that wording and the same model behaves differently.
How it works
Every prompt, word for word
Settings lists all five and when each was last edited. Open one to read the full wording and where it runs.
Your wording from the next request
Change a line and press Save. The assistant follows the new wording from then on, for everyone in the workspace.
Reset to the shipped wording
One click puts the default back in the editor. Nothing changes for the workspace until you press Save.
The five prompts
Each one is edited, saved and reset on its own.
What you get
In the product today
One per workspace
Everyone in a workspace gets the same wording, and each workspace keeps its own.
Admins can save
Owners and admins hold settings:write by default. Other built-in roles can only read.
Untouched stays current
A prompt you never edit is not stored, so it runs the latest shipped wording.
One prompt acts
Only the chat prompt drives a model that can change your mailbox.
Good practice
Getting the most out of it
- 01
Small edits
Change one rule at a time, so you can tell which edit moved the answers.
- 02
Keep a copy
Paste your version somewhere safe before a reset, because earlier versions are not kept.
- 03
Test after saving
Run one routine request through the chat or the composer, since new wording applies from the next one.
Questions
Asked often
Keep going
Works well with
Start
Your domain,
your mail.
Point a domain at OpenEmail and read it in a mailbox built around it. The free plan covers one domain.
