About
GraphFlow started as an internal experiment: chat clients store text, but knowledge lives in the links between messages. We rebuilt the chat around those links.
Our first version was a single Neo4j query that walked a support branch after an outage. It answered in seconds what took an hour of scrolling. The rest of the product is that idea applied to everyday work: capture the conversation, keep the structure, let the graph do the remembering.
Today GraphFlow ships as a self-hosted stack: a Node.js API, PostgreSQL for durability, Neo4j for traversal, and a canvas frontend that renders the same data as nodes or as a mind map.
If a feature needs our servers to work, it is optional by design. Licensing and support are services, not dependencies.
We would rather store one extra edge than lose the reason two messages belong together.
Quotas, AI cost and licence windows are explicit in the product and in these pages, never in a footnote.
Weight results by branch depth, recency and author proximity rather than keyword frequency alone.
Multiple cursors and annotations on the same knowledge map, persisted per thread.
Slack and Zendesk importers that preserve reply structure, so your history arrives as a graph.
Device pairing with end-to-end key exchange for regulated environments.
graphflow.intranetos.com