research-hub-multi-ai

A multi-AI coordination router generates a .coord/multi_ai_plan.md file with plan_id, goal, success_criteria, and per-task briefs, enabling independent execution by leaf agents like Codex or Claude.

A multi-AI router plan is needed when a single round of work requires two or more delegates. If only a single delegate is needed, no router plan is produced and requests route directly to the appropriate leaf skill.

Structure a coordination plan by writing a .coord/multi_ai_plan.md containing plan_id, created_utc, goal, success_criteria, and per-task briefs with their own brief_path, allowing leaves to emit result.json for reconciliation.

Yes, the multi-AI router supports multi-step sequences with cross-artifact coordination, enabling complex pipelines that connect Zotero, Obsidian, and NotebookLM workflows across different AI agents.

Results are reconciled through a central flow where each leaf agent emits a result.json file, allowing the coordination plan to verify outputs against the success_criteria defined in the .coord/multi_ai_plan.md.

The router is limited to scenarios requiring two or more delegates; it will not generate a coordination plan for single-delegate tasks, routing them directly instead, and requires leaf agents to independently emit result.json for reconciliation.