Knowledge-graph chat · self-hosted · graphflow.intranetos.com
Community is free forever; paid plans start with a 14-day trial. GraphFlow turns every message into a node in a live knowledge graph. Replies become branches, topics become hubs, and the AI summarises a long branch into two sentences the moment it grows past ten nodes.
Deployed on your own server in one command · Support support@intranetos.com · +2 01001945390
Every team already has the answers - buried in channels, tickets and old threads. Search returns keyword matches, not context. Nobody can see which decision produced which outcome, or who is still blocked.
GraphFlow stores the relationships, not just the text: who replied to whom, which follow-up belongs to which incident, which three messages are about the same topic.
Each message is a node with a stable id, an author, a depth and edges to its parent, its thread and similar content.
A 40-message incident becomes one branch you can collapse, expand and walk again next quarter.
Past ten nodes in one branch, the engine writes a two-sentence summary and attaches it as a node in the graph.
Your server, your PostgreSQL, your Neo4j, your keys. Message bodies are AES-256-GCM encrypted at rest.
Four moves turn a single reply into a graph you can walk, collapse and summarise - each one visible in the workspace while it happens.
The client writes to IndexedDB first and queues the operation, so nothing is lost when the connection drops.
client · local-first queueEvery write passes through the graph mapping router: the row lands in PostgreSQL and the node with its edges is mirrored into Neo4j.
api · dual writeNode view renders the whole thread; mind-map view draws users as sources and messages as bubbles with directed replies you can drag and click.
canvas · d3 force layoutAbove ten nodes in one branch the engine asks the model for a two-sentence summary, stores it as a node and broadcasts it to every open client.
ai · summarised branchGroup an incident, link the root cause, summarise the fix, and find the same pattern six months later.
One thread per topic, tags as hubs, similar-content edges surfacing ideas nobody linked by hand.
Give each client their own tenant, branding, storage quota and licence - from one deployment.
Buy a licence, receive a signed key, run it on your infrastructure. Plans differ by seats and storage; the owner dashboard issues, renews and revokes keys, and each plan carries its own storage quota enforced by the API.
graphflow.intranetos.com