Creating an Agent
AI funding and usage
AI included uses REDA credits. Bring your own AI uses your organisation’s provider account, billed directly by that provider. The licence selects one funding model for the organisation. In either model, REDA AI governs configured tools and actions in Salesforce. Set up your licence and AI accounts.
With AI included, select the agent’s Processing Tier. With Bring your own AI, select AI Connection and Model in the agent’s AI Account section, or inherit the organisation default. Different agents can use different supported providers and models. Prompts, tools, record context and channel permissions still define each agent’s job. Configure the agent’s AI account.
The New Agent Workflow
Creating a new agent in REDA AI is straightforward. You'll start with the essentials (name, channel, persona) and then build out the configuration by adding prompts, tools, and context fields. The wizard guides you through the process, and you can always come back to refine your agent's settings.
Step-by-Step: Creating a New Agent
-
Navigate to REDA AI App
Open the REDA AI app from the Salesforce app launcher.
-
Go to Agent Management Tab
Click the Agent Management tab to view all agents in your org.
-
Click New Agent Button
In the toolbar, click the "New Agent" button to start the creation wizard.
-
Fill in Display Name
Enter the name users will see in the chat UI (e.g., "Sales Assistant", "Customer Support Bot"). This field is required.
-
Enter Agent ID
Choose a short, unique identifier for the agent (up to 80 characters — keep it lowercase and memorable, e.g. "sales_bot"). It is used in Flow integrations and API calls, and it is required.
-
Select Channel
Choose where the agent will be deployed: Internal (Salesforce users), External Widget (website visitors), Community (portal users), or API (headless). This field is required.
-
Enter Persona
Write the opening system prompt that defines your agent's role, tone, and purpose (up to 5000 characters). Example: "You are a helpful sales assistant focused on closing deals..."
-
Save the Agent
Click Save to create the Agent Definition record with the Display Name, Agent ID, and Channel you entered.
-
Configure Additional Settings
After creation, add prompts, context fields, tools, and other settings as needed.
Required vs. Optional Fields
| Field | Required? | Notes |
|---|---|---|
| Display Name | Yes | Name shown to users. Can contain spaces and special characters. |
| Channel | No | Determines where the agent can be deployed and how callers authenticate. Salesforce does not block a blank Channel, and blank behaves as Internal at run time — so set it deliberately rather than relying on the default. |
| Persona | No | Recommended. Defines the agent's identity and purpose. Can be added later. |
| Agent ID | Yes | Required and unique (up to 80 characters). Used in Flow integrations and API calls. Keep it short, lowercase, and memorable. |
| Is Active | No | Defaults to active — untick it before saving if you are not ready to go live. If your plan's active-agent limit is already reached, saving an active agent is blocked until another agent is deactivated. |
Customizing the Agent ID
Every agent needs an Agent ID — a unique identifier you choose when creating it. This ID is used in Flow integrations and API calls. For example, if your agent is called "Sales Assistant", you might set the Agent ID to "sales_bot" or "sa_001".
Keep the Agent ID short, alphanumeric, and lowercase for ease of use in integrations. Once set, the Agent ID should not be changed frequently, as it may break existing Flow integrations or API references.
Cloning an Existing Agent
Instead of starting from scratch, you can clone an existing agent to use as a template. Cloning is useful when you want to create a similar agent for a different team or use case. For example, you might clone your Sales agent to create a separate agent for a different region, then customize the persona and tools for that region.
- Open the Agent Definition record you want to clone.
- Click the Salesforce standard "Clone" button (or select Clone from the dropdown).
- Update the Display Name and Agent ID for the new agent.
- Click Save to create the cloned agent.
- Re-create the related configuration. The standard Clone button copies only the Agent Definition record itself — Custom Prompts, Context Fields, tool assignments, and Suggestion Chips are not copied. Add them to the new agent (Build Agent with AI can rebuild prompts and context fields quickly, or use Data Loader to copy the related records).
What Happens After Creation?
Once you've created an agent, it is active by default — untick Is Active at creation if you would rather configure it first. You can now:
- Use "Build Agent with AI" to get AI-powered suggestions for prompts and context fields
- Add Custom Prompts to refine the agent's behavior and tone
- Add Context Fields to load Salesforce record data into the agent's context
- Assign Tools to enable the agent to query and modify Salesforce data
- Add Suggestion Chips for quick-action buttons in the chat UI
- Test the agent in a sandbox environment
- Activate the agent and deploy it to production
Best Practices for New Agents
Always test your agent in a sandbox environment first. Use the chat widget to verify the agent's responses, tool calls, and overall behavior. Make adjustments to prompts and settings based on testing, then deploy to production via Data Loader or Salesforce CLI.
Use "Build Agent with AI" immediately after creation. This feature analyzes your agent's configuration and suggests improvements to the persona, prompts, and context fields. It can save you hours of manual configuration work.
Start with the essentials. You don't need to assign every tool or context field on day one. Begin with the core tools and fields your agent needs, test it, then expand based on user feedback and usage patterns.
Name your agents clearly. Use descriptive display names that make it obvious what the agent does. Avoid generic names like "Bot" or "Assistant" — instead, use "Sales Pipeline Assistant" or "Customer Support Bot".
Using Build Agent with AI
After creating an agent, you'll see a "Build Agent with AI" button on the Agent Definition record. This feature uses AI to analyze your org's Salesforce schema and your agent's purpose, then recommends:
- Persona improvements to better define the agent's role and tone
- Custom Prompts to add behavioral instructions
- Context Fields to automatically load relevant Salesforce data
- Suggestion Chips for quick-action buttons
You review the suggestions and decide which ones to apply. This AI-powered approach gives you a production-ready agent in minutes, instead of hours of manual configuration.
Deploying to Production
Once your agent is tested and ready, deploy it from sandbox to production using standard Salesforce tools:
- Data Loader — Export the Agent Definition and related records from sandbox, then import them into production.
- Salesforce CLI — Use the CLI to pull changes from sandbox and deploy them to production.
- Change Sets — If your org has change set deployment enabled, bundle your agent's records in a change set and deploy to production.
Frequently Asked Questions
Yes, you can edit the Agent ID after creation. However, if the ID is already referenced in Flow integrations or API calls, changing it will break those integrations. Update all references in Flows and code before changing the ID.
Channel is a property of the Agent Definition that determines where the agent can be deployed (Internal, External Widget, Community, or API). After creating an agent with a channel, you then deploy it to specific pages or applications using Lightning App Builder or other deployment mechanisms.
Yes, the Persona field is optional during creation. However, we recommend always providing a clear persona. If you skip it, the agent will use a default system prompt, which may not be ideal for your use case. You can always add or update the persona later.
Creating an agent with basic configuration (name, channel, persona) takes less than 5 minutes. With full configuration (prompts, tools, context fields, testing) expect 30-60 minutes depending on complexity. Using "Build Agent with AI" significantly speeds up the process.
Next Steps
After creating an agent, explore agent configuration options to fine-tune its behavior, or learn about Build Agent with AI to get suggestions automatically.