Capacities
Planning
When Planning is enabled, the Agent automatically generates an execution plan at the start of each task. This plan evaluates available tools, connected MCPs, and steps needed to achieve the goal.
Ideal for:
Agents that interact with multiple APIs or Workflows.
Problem-solving tasks that require structured execution.
Scenarios where the Agent needs to reason before acting.
Supervision
Supervision adds a self-review mechanism to the reasoning loop. When enabled, the Agent:
Reviews the plan and actions performed.
Evaluates outcomes.
Adjusts future reasoning steps to improve accuracy.
This capacity is especially effective for long, multi-turn tasks where iterative optimization is valuable (e.g., research, summarization, debugging).
Last updated