notion-blackboard

A Notion-first multi-agent workflow where Notion is the shared operating surface, not just the place where an AI dumps text.

Notion Blackboard turns a Notion workspace into a coordination layer for multiple agents. You write an objective in Notion, a manager agent decomposes it into tasks, a worker agent produces intermediate notes, a reviewer agent validates the work, and the system publishes one clean final report back into Notion.

> Objectives in, final reports out. The rest is the engine room.

Most AI-to-Notion demos stop at "generate text and save it to a page." This project explores a different pattern: using Notion itself as the shared blackboard between agents.

The user sees a simple product surface:

• Start Here: explains the workspace
• Objectives: where new missions are entered
• Final Reports: where the final human-facing deliverables land

The internal workflow remains inspectable:

• Task Queue: manager-created work items
• Results: worker-produced intermediate notes
• Audit Log: review and orchestration trace
• Agent Registry: active agents and model visibility

Human
  |
  v
Objectives (Notion)
  |
  v
Manager agent -> Task Queue
  |
  v
Worker agent -> Results
  |
  v
Reviewer agent -> Audit Log
  |
  v
Final Reports (Notion)

Each agent reads from and writes to Notion. The agents do not need to talk directly to each other; Notion carries the workflow state.

• Notion-first intake through an Objectives database
• Multi-agent pipeline: manager, worker, reviewer
• Final human-facing deliverable in Final Reports
• Internal traceability through task, result, audit, and registry databases
• Bootstrap command to create the Notion workspace structure
• Doctor command to validate database schemas before a live run
• Markdown-to-Notion block rendering for readable result pages
• Local artifacts for demos and writeups in docs/
• Dry-run mode for local