> ## Documentation Index
> Fetch the complete documentation index at: https://docs.agents.labs.bandwidth.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Edit a workflow

> Build conversation steps, connect pathways, and test the complete flow.

Open a Workflow agent from **Agents**. You need Editor access or above to save changes.

## Configure conversation steps

Select a node to edit its prompt and settings. Give each step one purpose, such as greeting the caller, collecting details, or confirming a result. Use the [Global node](/voice-agent/global) for shared instructions and enable **Add Global Prompt** on steps that should use them.

## Connect pathways

Create a pathway between two nodes and describe the condition for taking it. Use distinct conditions when a step has several destinations. Add transition speech only when it helps the caller understand the move.

## Control the conversation

* **Allow Interruption** controls whether the caller can interrupt that step.
* **Wait for user response** controls whether the agent waits for input.
* **Detect Voicemail**, where available, changes how the agent handles an answering machine.
* **Variable extraction** captures structured values for the run and subsequent integrations.

Review the saved settings on each node; templates can choose different defaults.

## Connect actions and results

Attach [HTTP tools](/voice-agent/tools/http-api) or [MCP tools](/voice-agent/tools/mcp-tool) to the steps that need them. Add a [Webhook](/voice-agent/webhook) to deliver results after a call and [QA](/voice-agent/qa) to evaluate it.

## Verify the flow

Save your draft and run a browser test for every important path, including missing data and tool failures. Review the conversation and publish when ready. See [drafts and publishing](/voice-agent/publish).