Unlike external integrations that specialize in outside software, internal AI agents are designed to handle your native Dart workflows. You control exactly how they behave by setting up custom guidelines, choosing their underlying AI model (Auto/Dart AI, Claude, ChatGPT, or Gemini), and defining their automated workflow status changes.
Creating your own agent
Navigate to Agents in the left navigation bar and select Create a custom agent
Give your agent a specific Name and assign an Icon so your team can easily recognize it
In the text box, describe exactly what this agent should do (see tips on writing guidelines below)
Select the AI model that will power the agent from the dropdown (Auto, Claude, ChatGPT, or Gemini)
Configure how the agent interacts with your workflow. You can define specific rules such as shown above.
How to assign work to internal agents
Internal agents are seamlessly integrated into your standard Dart workflow. To trigger an internal agent, simply assign it to a task. As soon as it is added as the assignee, the agent will read the task, follow its custom guidelines, and execute the work while automatically updating the task statuses based on your workflow settings.
Writing effective custom guidelines
Your agent's guidelines are its rulebook. To get the best results, be highly specific about the steps the agent should take when it receives an assignment.
Examples of custom agent guidelines:
Bug & issue triage agent
"When assigned a new bug report, scan the description for clear reproduction steps and error logs. Check for duplicate active tasks in Dart. If this is a critical backend issue, prioritize it as 'High' and suggest assigning it to Carter or Nick. Finally, add a concise summary comment for the engineering team."
Integration & payload parsing agent
"When assigned a task created via n8n or Zapier webhooks, extract the raw JSON data from the description and convert it into a clean, readable Markdown table. Identify any missing user parameters, assign a size estimate based on the payload complexity, and route the cleanly formatted task to the technical support queue."
Operations QA & review agent
"When assigned a task for final review, verify that all subtasks are checked off and that the original deliverables meet the requirements. If anything is missing, change the status to 'Blocked' and leave a comment tagging the Area Lead (AL) with the specific missing details. If everything is fully complete, mark the task as 'Done'."
Standup prep agent
"Every morning, look at tasks that were updated yesterday, summarize progress, flag anything that's been stuck for more than 3 days, and create a summary doc."
Triage agent
"When a new bug report comes in, check if it has reproduction steps. If severity isn't set, estimate it based on the description. Assign it to the right team based on the component mentioned."
Onboarding agent
"When someone is added to the project, create a set of onboarding tasks assigned to them from this template, and create a welcome doc with links to key resources."
Sprint cleanup agent
"At the end of a sprint, move incomplete tasks to the next sprint, add a note about why they rolled over, and create a summary doc of what shipped."