Product Model
Breakdown is a node-based reasoning canvas. Users structure source material and analysis steps as a directed graph, then run nodes in context.
Core Idea
Instead of leaving research scattered across chat threads and documents, Breakdown turns a reasoning process into connected nodes. Source nodes hold material, analysis nodes transform upstream context, and edges describe how results should flow.
Why A Graph
Many decisions depend on multiple assumptions, evidence streams, and sub-conclusions. A graph makes those dependencies visible, lets users refresh source material, and helps downstream analysis stay connected to what changed upstream.
| Node Family | Role |
|---|---|
| Sources | External material such as Google Drive files, web pages, pasted text, or imported data. |
| Analysis steps | Prompted reasoning units that consume upstream outputs and produce generated results. |
| Compositions | Higher-level conclusions or decisions assembled from upstream analysis. |
Edge Semantics
Edges are more than lines on the canvas. Their labels tell the runner how to frame upstream outputs when building the prompt context for downstream nodes.
| Edge Type | Meaning |
|---|---|
| supports | An upstream result strengthens the downstream analysis. |
| contradicts | An upstream result weakens or challenges the downstream analysis. |
| depends_on | A downstream step requires the upstream result to remain valid. |
| assumes | A downstream step is predicated on an upstream assumption. |
| inputs_to | An upstream output is direct context for downstream work. |
External Agents
Breakdown can also act as a headless graph layer. External AI clients can connect over MCP, inspect graph structure, run workflows, and submit results while keeping their own connector and model environment.
See MCP Access for connection details.