Edit Assistant

Change an assistant's settings, such as its instructions, knowledge, tools, safety rules, or model, and then publish the changes to the people who use it. You always edit a private copy first, so you can try changes without affecting anyone.

Open the Assistants section. If you're not sure how, see Add Assistant.

How editing works

Every assistant has two versions:

Production is the live assistant that people and integrations use. You can't change it directly.

Staging is your private working copy. Every edit you apply goes to Staging, and a counter next to Deploy shows how many changes are waiting.

When you're happy with your changes, you deploy them, which copies Staging to Production. The first deployment also turns a Draft assistant Live. Every deployment is recorded, and you can restore an earlier one. See Staging & Publishing.

Use cases

Tighten answers: People say an assistant drifts off topic. Add a custom instruction such as "Answer only from the travel policy," test a few questions in Staging, then deploy.

Add knowledge: A new policy document is published. Add it to the assistant's corpus in Staging, check that the assistant uses it, and deploy.

Switch models: Try a different model in Staging to compare answer quality or speed before anyone else sees the change.

1. Open your assistant

Search for your assistant, then click the chat icon (Chat with assistant) in its Actions column.

2. Switch to Staging

At the top right, click Staging. The banner changes to Edits are private until deployed. If the assistant already opens in Staging, you can skip this step.

3. Change the settings

In the Configuration panel on the right, open the section you want, make your changes, and click Apply Changes. The message Staging assistant updated successfully! appears and the change counter goes up.

Section	What you can change
Assistant Tools	Turn tools on or off. Each tool is marked Internal (built into Ejento AI, for example Retriever Tool and Attachment Tool) or External (connected to another application). See Tools.
Corpus	Add or manage the documents the assistant searches. See Corpus Overview.
Guardrails	Turn on safety filters: Topical Control (keeps conversations to allowed topics), Ethical Moderation (keeps responses ethically sound), Jailbreak Detection (blocks attempts to trick the assistant into ignoring its rules), Hallucination (reduces made-up answers), and PII Redaction (hides personal information). See Understanding Guardrails.
Customize	Assistant Role (the assistant's main role), Custom Instructions (extra rules that always take priority over standard instructions; click + to add more), Conversation Starters (suggested questions shown to users), and Tags.
Advanced	Assistant Default Model (the language model that writes responses), Reasoning Pattern (see Reasoning Patterns), Visibility (Private or Public; see Assistant Access Control), Publish to Coworkers, Call (voice calls; see Voice Calling with Assistants), and Enable Embed Link (see Embed Assistant).

4. Deploy your changes

Click Deploy. The Deploy to Production window lists each setting that changed, with its old and new value. Optionally add a Description of what changed and why (up to 255 characters), then click Deploy Changes.

The message Successfully deployed to production appears, and the assistant switches to Production with your changes live. If you see Nothing to deploy, Staging already matches Production.

The first deployment also changes the assistant's status from Drafted to Live. To review past deployments or restore an earlier one, see Staging & Publishing.

How editing works#

Use cases#

1. Open your assistant#

2. Switch to Staging#

3. Change the settings#

4. Deploy your changes#

How editing works

Use cases

1. Open your assistant

2. Switch to Staging

3. Change the settings

4. Deploy your changes