Using AI agents
After your organization deploys AI agents, you can interact with them in a conversational user interface. Use natural language prompts to initiate an agent, similar to how you would ask a colleague for help. If you need help getting started with prompts, select a conversation starter tile available in the Chat screen.
In addition to using AI agents in the conversational interface, you can embed agents you create in Agentstudio directly into your own applications using the Boomi Agentstudio Embed Kit. Get started with the GitHub repository and explore the Boomi Embedded documentation.
Prerequisites
You must have the Agent User role or a custom role with Agent Garden and Agent View privileges to chat with deployed agents.
By default, all Platform Standard role users have the Agent Garden User role, which allows them to interact with deployed agents in the Chat window. All Platform administrators have the Agent Garden administrator role.
Agent conversation
In the Boomi Platform click the AI icon and navigate to Agent Garden > Chat screen.
Select a runtime cloud from the region selector. Agent cards appear for each agent with an active deployment in the selected runtime. Each card displays the agent's provider, runtime cloud, and package version.
Select an agent card to start a conversation. Once in the conversation, you can view the runtime cloud, environment, and package version by hovering over the agent title.
If the agent has Extended Thinking enabled and no guardrails set, you'll see its plan stream in as it's generated, instead of a static "Thinking..." message.
Boomi GPT is only available on the USA East Agent Garden Cloud 01 runtime.
Viewing session logs of Boomi Agent Garden agents
View session logs containing details of deployed Boomi garden agents behaviour, performance, and troubleshooting issues. You can click on View logs icon directly from the Chat screen. Read Tracing sessions of Boomi Agent Garden agents to learn more.
Favorite agents
Agent Garden makes it easy for you to interact with your favorite agents. Select Add to Favorite to add deployed, active agents to your favorites list.
Clearing conversation memory
As you converse, the agent retains a memory of your inputs and information from any tools it uses. This memory helps the agent reason more effectively, understand context, and respond faster to prompts to achieve tasks and the overall goal. With Extended Thinking on, the agent also retains tool response data outside the active context, so it can recall earlier results instead of calling the same tool again. You can clear this memory at any time by resetting the conversation.
Click Start New Conversation to clear the conversation memory, which removes context and previous information when an agent performs tasks. Starting a new conversation does not carry over memory from earlier conversations with the same agent, even if those conversations remain in your history. Resetting the conversation does not delete the conversation history. The Agent Garden retains history until you choose to delete it.
Reviewing and deleting conversation history
You can refer back to past conversations with an AI agent to review its responses. Click the Conversation History button to see a summary of each conversation along with the start time.
Click the Delete icon to permanently remove the chat from your conversation history. Deleting a conversation also permanently deletes any memory the agent retained for that session, including tool response data.
Invoking agents using the Agent Garden REST APIs
Beyond the conversational chat interface, agents can be invoked via REST APIs. This enables:
- Integrating agents into Boomi Integration processes (via Agent step)
- Building custom applications that call agents
- Automating multi-agent orchestrations
- Creating agent-driven workflows in third-party systems
Refer to Invoking agents via API for request/response formats and examples.
