# Workflow A **workflow** is a defined sequence of steps that an AI [agent](agent) follows to complete a task. Workflows specify the order of actions, decision points, tool calls, and [human-in-the-loop](human-in-the-loop) touchpoints that guide an agent from a starting state to a desired outcome. ## Workflows vs. Agentic Workflows The term *workflow* is broad. In the context of AI agents, an [agentic workflow](agentic-workflow) is a workflow where the agent has autonomy to decide its own path rather than following a rigid script. | Type | Agent Freedom | Human Involvement | |------|--------------|------------------| | Scripted workflow | Follows fixed steps | Defined checkpoints | | Agentic workflow | Adapts based on results | Dynamic oversight | ## Components of an Agent Workflow A typical agent workflow includes: 1. **Trigger** — An event or instruction that starts the workflow 2. **Planning** — The agent determines the approach 3. **Execution** — The agent runs tools and produces outputs 4. **Checkpoints** — Points where the agent pauses for human [approval](approval) or input 5. **Completion** — The agent signals the workflow is done ## Designing Workflows with AgentRQ AgentRQ enables you to build [human-in-the-loop](human-in-the-loop) workflows for [Claude Code](claude-code) agents. A well-designed AgentRQ workflow looks like: 1. Agent receives a task (e.g., "implement OAuth login") 2. Agent calls `getWorkspace` to understand project context 3. Agent creates a [task](task) via `createTask` with status `ongoing` 4. Agent works autonomously — reading code, writing implementation 5. Agent sends [notification](notification) via `reply` when it reaches a decision point 6. Human reviews and responds 7. Agent continues to completion 8. Agent calls `updateTaskStatus` to mark the task `completed` ## Related Terms - [Agentic Workflow](agentic-workflow) - [Task](task) - [Human-in-the-Loop](human-in-the-loop) - [Approval](approval) - [Agent](agent)