stream-chain

Multi-agent orchestration streams each agent's output directly into the next step's input, enabling cohesive sequential processing. You can execute custom chains or predefined pipelines using the claude-flow CLI to coordinate agents and preserve context across steps.

Predefined pipelines are battle-tested workflows for common tasks like codebase analysis. You can run them via the claude-flow CLI using the pipeline command, which applies a standardized sequence of prompts to accelerate development and produce coherent results.

Custom chains execute user-defined prompt sequences with full control over each step. Each step's input builds on the previous output, allowing you to perform iterative refinement and reproducible software development tasks by streaming context across steps.

Multi-agent streaming supports optional flags like --verbose and --timeout through the claude-flow CLI. These flags allow you to monitor pipeline execution details and manage execution limits to prevent hanging during complex orchestration.

Use predefined pipelines for common, standardized data processing tasks to leverage battle-tested workflows and accelerate development. Choose custom chains when you need full control over user-defined prompt sequences and iterative refinement for specialized processing requirements.

Context flow preserves memory across steps by streaming each agent's output into the next. This mechanism ensures iterative refinement and reproducibility, allowing downstream agents to build directly on the context generated by previous steps.