Creating Agents

This guide walks you through creating a new AI agent from scratch, covering configuration options and best practices.

Prerequisites

Access to Agent Studio
Understanding of your use case and goals
Knowledge of what tools the agent will need

Step 1: Start a New Agent

Navigate to Agent Studio from the main navigation
Click Agents in the sidebar
Click Create Agent or the + button

You’ll see the agent configuration form.

Step 2: Basic Information

Agent Name

Choose a descriptive name that indicates the agent’s purpose:

Good Names	Poor Names
”Account Health Analyzer"	"Agent 1"
"QBR Prep Assistant"	"Helper"
"Renewal Risk Monitor"	"AI Bot”

Names should help users quickly understand what the agent does without reading the full description.

Step 3: Write Instructions

Instructions (the system prompt) define how your agent behaves. This is the most important configuration.

Get help from Stella

Stella can draft Instructions for a new agent or review the Instructions of an existing one. Use the button next to the Instructions label:

Help me write appears when the Instructions field is empty.
Review this prompt appears when the field already has content.

Either button opens a modal.

Help me write
Review this prompt

Answer three questions:

The job: what the agent should produce. A sentence is enough. For example, “a renewal-risk brief for one account: the open risks with evidence, and three next steps.”
Who reads the result? (optional): for example, “the account’s CSM.”
When does it run?: Only when I ask or On a schedule.

Click Write the Instructions. Stella works out which toolkits and data the agent needs, checks they exist in your workspace, and writes and tests the Instructions.

Stella returns a proposed configuration card with:

The Instructions: click Read the draft to see new Instructions, or See what changed to see a diff of your existing Instructions.
Toolkits: the toolkits Stella picked for the agent.
Decided for you: choices Stella made on your behalf. Change any of them and ask again.
Still open: questions Stella couldn’t resolve.
Schedule (optional): the cadence the Instructions were written for.

Click Apply to this agent to fill in the builder. Apply does the following:

Replaces the Instructions with the proposed version.
Adds any proposed toolkits that aren’t already enabled.
For a new agent, fills in a suggested name and description if those fields are empty.

Apply doesn’t save the agent or set a schedule. Review the changes in the builder, then click Save.

Structure Your Instructions

# Role
[Who the agent is and what it specializes in]

# Primary Tasks
[What the agent should do when asked]

# Guidelines
[How the agent should approach tasks]

# Output Format
[How responses should be structured]

# Constraints
[What the agent should avoid]

Example: Customer Health Agent

# Role
You are a customer success analyst specializing in account health assessment and early warning detection.

# Primary Tasks
When analyzing an account:
1. Review health score trends over the past 90 days
2. Analyze engagement patterns from activities
3. Identify key contacts and their recent involvement
4. Compare against similar successful accounts
5. Generate specific, actionable recommendations

# Guidelines
- Base all conclusions on data, not assumptions
- Prioritize actionable insights over general observations
- Consider both quantitative metrics and qualitative signals
- Explain your reasoning clearly

# Output Format
Structure your analysis as:

## Health Assessment
[Overall assessment with health score context]

## Key Findings
- [Finding 1 with data support]
- [Finding 2 with data support]

## Risk Factors
- [Risk 1]: [Mitigation suggestion]

## Recommended Actions
1. [Specific action with expected outcome]
2. [Specific action with expected outcome]

# Constraints
- Never fabricate data or statistics
- Don't make promises about outcomes
- If data is insufficient, say so explicitly

Step 4: Select the Model

Choose the AI model based on your needs:

Model Options

Model	Speed	Capability	Cost	Best For
Gemini Auto	Varies	Varies	Optimized	Most use cases
Gemini 3 Pro	Slower	Highest	Higher	Complex analysis, critical tasks
Gemini 3 Flash	Fastest	Good	Lower	Quick queries, simple tasks
Gemini 2.5 Flash	Fast	Very Good	Moderate	Balanced needs

Recommendations

Start with Gemini Auto - Let the system optimize model selection
Use Pro for complex research, multi-step analysis, or critical decisions
Use Flash for quick lookups, simple formatting, or high-volume tasks

Step 5: Configure Reasoning Level

Reasoning level controls how much the agent “thinks” before responding:

Options

Level	Behavior	Use When
Auto	Adapts to task complexity	You’re unsure what level is needed
Low	Quick, direct responses	Simple questions, factual lookups
Medium	Moderate analysis	Standard analysis, most tasks
High	Deep reasoning	Complex problems, nuanced analysis

Higher reasoning = better analysis but slower responses and higher cost. Match the level to your actual needs.

Step 6: Enable Toolkits

Toolkits give your agent capabilities. Enable only what’s needed.

Available Toolkits

Toolkit	Capabilities	Enable When
Customer Intelligence	Read accounts, contacts, activities, meeting transcripts, upcoming meetings, goals, opportunities, support tickets, tasks, projects, product usage, and health changes	Agent answers questions about your customers
Data Analysis	Run Python code for accurate counting, aggregation, and data calculations	Agent needs precise arithmetic or data analysis
Web Research	Search the web for information about companies, people, and news	Agent needs outside context, such as company news
Custom Field Tools	Read and update custom field values on accounts, contacts, opportunities, tasks, and people. Updates ask for confirmation.	Agent records its findings in fields
Document Tools	Read Statisfy documents, and create or update them. Saving asks for confirmation.	Agent writes or edits documents
Digital Workbench	Draft emails, Slack or Microsoft Teams messages to a customer’s channel, and documents that you review, edit, and send yourself	Agent prepares outreach for a person to review
Slack	Send messages to Slack channels and list available channels. Sending asks for confirmation. Available once Slack is connected.	Agent posts updates or alerts in Slack
Microsoft Teams	Send messages to Teams channels and list available channels. Sending asks for confirmation. Available once Microsoft Teams is connected.	Agent posts updates or alerts in Teams
MCP servers	Tools from the third-party MCP servers your organization has connected. Each connected server appears as its own toolkit.	Agent works in another tool through its MCP server
Zendesk, Jira, Salesforce CPQ, Notion (API Tools)	Read and write records in your connected integrations. See API Tools toolkits.	Agent works directly in a support, issue-tracking, quoting, or docs tool

API Tools toolkits

API Tools toolkits let your agent read and write directly in Zendesk, Jira, Salesforce CPQ, and Notion. Each toolkit uses your organization’s existing connection to that integration, so you don’t need to authorize anything new. In the Add Tools picker, API Tools toolkits appear in their own API Tools section, separate from the MCP Servers section. A vendor’s toolkit appears only when that integration is connected for your organization. To connect one, see the Zendesk, Jira, Salesforce, or Notion integration guide.

Toolkit	Read actions	Write actions
Zendesk (19 actions)	Search and get tickets, list ticket comments, search and get users, search and get organizations, list an organization’s tickets, list macros, list CSAT ratings	Create and update tickets, add a public reply or internal note, merge tickets, add and remove ticket tags, create or update users and organizations, apply a macro to a ticket
Jira (14 actions)	Search issues with JQL, get an issue, list comments, list available workflow transitions, list and get projects, search users	Create, update, and delete issues, add comments, transition issues (for example, to Done), assign issues, log work
Salesforce CPQ (17 actions)	Find and read quotes, quote lines, and products; run read-only Salesforce queries	Create and update quotes and quote lines, and take a quote through approval, order form, and signature
Notion (9 actions)	Search pages and databases, get a page, read page blocks, query database rows (data sources), list comments	Create and update pages, append blocks to a page, add comments

Approval defaults

Read actions run automatically.
Write actions ask for confirmation before they run, unless you configure the agent otherwise.

To change this for one agent, open the toolkit in the Add Tools picker. Each action, and each group of actions (Read-only tools, Write/Delete tools, and so on), has an Always allow / Needs approval dropdown; a group shows Custom when its actions differ. The same dropdowns appear for toolkits in the MCP Servers section. These settings belong to this agent only and override the server’s approval policy on the Integrations page, so two agents can treat the same tool differently. You can include or exclude individual actions for each agent. Enable only the actions the agent needs. For example, a ticket triage agent might need Zendesk search and comment actions but not merge or user updates.

Requirements

Salesforce CPQ: Quote and quote line actions need the Salesforce CPQ managed package installed in your Salesforce org.
Salesforce CPQ: Some quote actions (such as cloning a quote or generating its order form) work only when Salesforce is connected through the Salesforce (Direct) card on the Integrations page. If your org connects through the Salesforce card, those actions don’t work. See Salesforce for the two connection options.
Salesforce CPQ: Sending an order form for signature needs an e-signature app (such as Adobe Acrobat Sign or Docusign) installed in your Salesforce org, with a signature template set up for quote order forms.
Zendesk: Write actions need a Zendesk connection that granted write scope.

API Tools toolkits are separate from third-party MCP servers. MCP servers are connected under Integrations → MCP Servers. API Tools toolkits use the integrations you’ve already connected.

When you chat with an agent that has the Customer Intelligence toolkit, the agent can also run read-only SQL queries over Statisfy’s reporting views. There’s nothing to enable or configure. The agent uses these queries when the other data tools only return current state or recent data — for example, an account’s full health history and time spent in each state, product-usage rollups across many accounts, medians and percentiles, top-N rankings, or custom cross-entity joins. The views include accounts and account owners, health history, product usage, activities, support tickets, surveys, tasks, projects, contacts, opportunities, custom fields, and custom objects. Each query must be a single SELECT statement, sees only your organization’s data, and returns at most 200 rows within a 15-second timeout.

Guidelines

Less is more

Each enabled toolkit adds complexity. An agent with 3 focused toolkits will often outperform one with 10 toolkits.

Match to purpose

A health analysis agent needs Customer Intelligence. It probably doesn’t need Digital Workbench or Slack.

Consider the user

Will users expect the agent to take actions (update fields, send messages) or just provide information?

Step 7: Set Scope

Define what data the agent can access:

Scope Options

Scope	Access	Use For
Global	All organization data	General-purpose agents
Account	Specific account context	Account-focused workflows

Account Scope

When using account scope:

Agent can only access data for the specified account
Useful for embedding agents in account-specific contexts
Provides data isolation and relevance

Step 8: Add Datasets (Optional)

Connect specific data sources for the agent to reference:

Knowledge bases - Internal documentation
Playbooks - Standard procedures
Templates - Response templates

This gives the agent access to your organization’s specific knowledge.

Step 9: Add a Presentation Template (Optional)

If your agent generates presentations, you can give it its own branded PowerPoint template:

In the Presentation Template section, click Upload Template
Select a .pptx file (PowerPoint format)
Once uploaded, the section shows Template uploaded with Replace and Remove buttons

Template Mode

After uploading, choose how the agent uses your template:

Mode	Behavior	Use When
Brand Guide (Creative)	AI creates new slides inspired by your template’s look & feel	You want on-brand decks with flexible layouts (default)
Strict Template	AI fills content into your template’s existing slide structure	Your deck must follow a fixed slide-by-slide format

An agent’s own template takes priority over your organization’s default presentation template. To set the organization-wide default instead, see Stella AI advanced features.

Step 10: Test Your Agent

Before deploying, test thoroughly.

Test Scenarios

Typical request
- Ask something your users commonly would
- Verify the response is helpful and accurate
Ambiguous request
- Ask something vague
- Does the agent ask clarifying questions or make reasonable assumptions?
Edge case
- Ask about an account with missing data
- How does the agent handle incomplete information?
Out of scope
- Ask something outside the agent’s purpose
- Does it gracefully redirect or explain limitations?

Example Test Prompts

Test 1: "Tell me about Acme Corp's health"
Expected: Detailed analysis with specific metrics

Test 2: "Acme Corp"
Expected: Asks what information is needed OR provides summary

Test 3: "What's the weather?"
Expected: Explains it can't help with that / redirects to purpose

Step 11: Deploy

Once tested:

Click Save to save your agent
Toggle Active to enable it
The agent is now available for use

Configuration Examples

Research-Focused Agent

Name: Account Research Assistant
Model: Gemini 3 Pro
Reasoning: High
Toolkits: Customer Intelligence, Web Research
Scope: Global

Instructions:
You conduct deep research on customer accounts. When asked about
an account, provide comprehensive analysis covering health metrics,
engagement trends, key contacts, and strategic recommendations.
Always cite specific data points and explain your reasoning.

Action-Oriented Agent

Name: Field Updater
Model: Gemini 2.5 Flash
Reasoning: Medium
Toolkits: Customer Intelligence, Custom Field Tools
Scope: Account

Instructions:
You help users keep account fields current. When asked to update a
field (for example, renewal risk or next step), confirm the account,
the field, and the new value, then make the update. Keep interactions
concise and efficient.

Communication Agent

Name: Outreach Drafter
Model: Gemini Auto
Reasoning: Medium
Toolkits: Customer Intelligence, Digital Workbench
Scope: Account

Instructions:
You draft personalized outreach messages. When asked to draft an
email or message, research the account and contacts first, then
create tailored content that references specific context. Always
output in a ready-to-send format.

Best Practices Summary

Start simple

Begin with minimal toolkits and instructions. Add complexity only when needed.

Be specific in instructions

Vague instructions lead to inconsistent results. Be explicit about what you want.

Test with real data

Sample data doesn’t catch real-world issues. Test with actual accounts and scenarios.

Iterate based on feedback

Your first version won’t be perfect. Collect feedback and refine instructions.

Monitor performance

Track how the agent is being used and whether responses are helpful.

Next Steps

Agent Best Practices

Advanced tips for agent optimization

Use Agents in Automations

Embed agents in workflow automations

​Creating Agents

​Prerequisites

​Step 1: Start a New Agent

​Step 2: Basic Information

​Agent Name

​Step 3: Write Instructions

​Get help from Stella

​Structure Your Instructions

​Example: Customer Health Agent

​Step 4: Select the Model

​Model Options

​Recommendations

​Step 5: Configure Reasoning Level

​Options

​Step 6: Enable Toolkits

​Available Toolkits

​API Tools toolkits

​Approval defaults

​Requirements

​Guidelines

​Step 7: Set Scope

​Scope Options

​Account Scope

​Step 8: Add Datasets (Optional)

​Step 9: Add a Presentation Template (Optional)

​Template Mode

​Step 10: Test Your Agent

​Test Scenarios

​Example Test Prompts

​Step 11: Deploy

​Configuration Examples

​Research-Focused Agent

​Action-Oriented Agent

​Communication Agent

​Best Practices Summary

​Next Steps

Agent Best Practices

Use Agents in Automations

Creating Agents

Prerequisites

Step 1: Start a New Agent

Step 2: Basic Information

Agent Name

Step 3: Write Instructions

Get help from Stella

Structure Your Instructions

Example: Customer Health Agent

Step 4: Select the Model

Model Options

Recommendations

Step 5: Configure Reasoning Level

Options

Step 6: Enable Toolkits

Available Toolkits

API Tools toolkits

Approval defaults

Requirements

Guidelines

Step 7: Set Scope

Scope Options

Account Scope

Step 8: Add Datasets (Optional)

Step 9: Add a Presentation Template (Optional)

Template Mode

Step 10: Test Your Agent

Test Scenarios

Example Test Prompts

Step 11: Deploy

Configuration Examples

Research-Focused Agent

Action-Oriented Agent

Communication Agent

Best Practices Summary

Next Steps