research-orchestrator

Multi-agent workflow orchestration coordinates specialized agents to handle distinct research stages—data extraction, statistical analysis, drafting, and review—delegating tasks and validating outputs at each step to produce a final reviewed article with structured artifacts.

Yes. Claude Code can orchestrate end-to-end research workflows by delegating PDF data extraction to a researcher agent, correlation analysis to technical analysis, article drafting to a copywriter, and quality review to an antagonist agent, producing validated JSON outputs and a final markdown article.

Workflow orchestration validates that intermediate artifacts exist and conform to schema—parsed_papers.json contains all extracted papers, correlation_analysis.json has valid coefficients with |r| ≤ 1—ensuring data integrity before proceeding to downstream stages like drafting and review.

MCP-aware orchestration secures file access to research papers and outputs by delegating named skill invocations to specialist agents, managing data flow between stages, and enforcing validation thresholds so each agent receives clean inputs and produces provenance-rich artifacts.

Delegate the review stage to a research-antagonist agent that evaluates draft articles against research outputs and flagged limitations, producing structured review feedback that flows back into the orchestration pipeline for final delivery of validated research deliverables.