Creating Agents

Learn how to create new agents in Cuneiform Chat.

Getting Started

Navigate to Agents in the sidebar
Click the Create Agent button in the top right
Fill in the form sections below

Basic Information

This section captures essential details about your agent.

Field	Required	Description
Name	Yes	A name for your agent (e.g., “Customer Support Agent”). Maximum 100 characters.
Description	No	A brief description of what this agent does. Maximum 500 characters.

Tips:

Choose a descriptive name that indicates the agent’s purpose
The description helps you and your team understand the agent’s role

Agent Persona

The agent persona defines how your agent behaves and responds through a system prompt.

Key Elements:

Platform Rules Box — Shows built-in safety rules that are automatically applied (click to view details)
Generate Button — Opens the Agent Persona Builder to help create your persona
Edit Button — Opens a panel to manually write or edit the persona

Platform Rules

These safety guardrails are automatically applied to all agents:

No fabrication — Won’t invent facts not in your content
Acknowledges gaps — Says “I don’t know” when appropriate
Resists speculation — Won’t guess at missing information
No professional impersonation — Won’t act as lawyers, doctors, etc.
Corrects false premises — Points out incorrect assumptions
Blocks harmful content — Refuses inappropriate requests

Platform rules ensure all agents behave reliably and safely. They cannot be disabled. See Agent Persona Builder for details.

Quick Persona Example


You are a helpful assistant for [Your Company].
Answer questions based on the provided documentation.
Be concise and friendly.

See Agent Personas for detailed guidance on writing effective personas.

Language Support

Your agent automatically detects the language of each user message and responds in the same language — no configuration needed. 176 languages are supported out of the box. See Agent Personas — Language Detection for details.

LLM Configuration

This section controls the AI model settings.

Default Mode:

By default, LLM settings use your organization’s tier defaults. You’ll see an info box showing “Using tier defaults”. This is recommended for most use cases.

Advanced Mode:

Toggle on Advanced to customize:

Setting	Description
Model	Choose from available AI models based on your plan
Temperature	Controls response creativity (lower = more consistent, higher = more creative)

Show Document References:

This toggle (always visible) controls whether the agent includes source citations in responses:

Off (default) — Responses don’t mention source documents
On — Responses include references to source documents (useful for legal/compliance)

Knowledge Access Control

Control what documents your agent can access when answering questions.

Mode	Behavior	When to Use
Exclude Mode	Agent accesses ALL documents EXCEPT those selected	Hide sensitive documents from the agent
Include Mode	Agent accesses ONLY selected items	Restrict agent to a specific set of knowledge

When using Include Mode:

Select specific folders the agent can access
Select specific tags the agent can access
Select individual documents the agent can access

The agent will only use information from the selected sources when responding.

Start with Exclude Mode (full knowledge access) unless you need to hide specific documents. You can always restrict access later.

Form Actions

At the bottom of the form:

Button	Action
Cancel	Discard changes and go back
Reset	Clear all fields and start over (shows confirmation dialog)
Create Agent	Save the agent and open its edit page

Your draft is automatically saved as you work. If you leave and come back, your progress will be restored.

After Creating

When you click Create Agent:

The agent is created with Testing status
You’re taken to the agent’s edit page
Test the agent using the Playground
Connect integrations when ready to go live
Change status to Active when ready

New agents start in Testing status. You can test in the Playground, but channel integrations are disabled until you activate the agent.

Managing Agents

From the Agents page, you can:

Edit — Click on any agent to modify its settings
Test — Click Test on an agent card to open the Playground
Change Status — Set to Testing, Active, Inactive, or Archived

See Configuration Options for details on all agent settings.