Overview
This node integrates with the Claude Code AI SDK to perform AI-powered coding tasks through conversational interactions. It allows users to send prompts or instructions to Claude Code and receive responses that can include code generation, file edits, command execution, and more, leveraging built-in tools.
The "Continue" operation specifically continues a previous conversation with Claude Code, enabling multi-turn interactions where context is preserved across exchanges. This is useful for iterative coding assistance, debugging sessions, or complex workflows requiring back-and-forth communication with the AI.
Practical examples:
- Continuing a coding session where you iteratively refine a function or script.
- Following up on a previous query to add features or fix bugs in generated code.
- Maintaining context while exploring different approaches to a programming problem.
Properties
| Name | Meaning |
|---|---|
| Prompt | The instruction or message sent to Claude Code to continue the conversation. Supports expressions to dynamically use data from previous nodes. |
| Model | Selects the Claude model to use: - Sonnet: Fast and efficient for most tasks. - Opus: More capable for complex tasks. |
| Max Turns | Maximum number of conversation turns (back-and-forth exchanges) allowed in this continued session. |
| Timeout | Maximum time in seconds to wait for the AI response before aborting the request. |
| Project Path | Directory path where Claude Code should run commands or access files. If empty, uses the current working directory. Useful for project-specific context. |
| Output Format | Determines the format of the output data: - Structured: Returns an object with messages, summary, result, and metrics. - Messages: Returns raw array of all exchanged messages. - Text: Only final result text. |
| Allowed Tools | Built-in tools Claude Code is permitted to use during execution, such as Bash commands, file editing, web search, etc. Multiple selections allowed. Defaults include web fetch, todo write, web search, exit plan mode, and task launching. |
| Additional Options | Collection of optional settings: - System Prompt: Extra context or instructions for Claude Code. - Require Permissions: Whether tool usage requires explicit permission. - Debug Mode: Enables debug logging. |
Output
The node outputs JSON data structured according to the selected output format:
Structured: An object containing:
messages: Array of all messages exchanged during the conversation.summary: Counts of user messages, assistant messages, tool uses, presence of results, and available tools.result: The final result text or error message.metrics: Performance data including duration (ms), number of turns, cost, and usage statistics.success: Boolean indicating if the operation succeeded.
Messages: Raw array of all messages exchanged, including user inputs, assistant replies, and tool usage events.
Text: Only the final result text from the assistant, along with success status, duration, and cost metadata.
The node pairs each output item with its corresponding input item index.
Dependencies
- Requires an API key credential for authenticating with the Claude Code service.
- Uses the official Claude Code SDK (
@anthropic-ai/claude-code) for interaction. - Node configuration may require setting appropriate timeout values and project paths depending on the environment.
- Optional debug logging can be enabled for troubleshooting.
Troubleshooting
- Empty Prompt Error: The prompt must not be empty; ensure the prompt property is set and contains meaningful text.
- Timeout Errors: If the operation times out, consider increasing the timeout value in seconds under Additional Options.
- Permission Denied for Tools: If tool usage fails due to permissions, enable "Require Permissions" or adjust allowed tools accordingly.
- API or Network Issues: Failures connecting to Claude Code service will throw errors; verify API credentials and network connectivity.
- Debugging: Enable debug mode to get detailed logs about the execution flow and received messages, which helps diagnose issues.