BixRouter

Most chat interfaces assume you know exactly where you’re going. Bix Router is built for the opposite situation — exploratory conversations where you want to follow multiple threads without sacrificing the ones you didn’t choose. The workflow runs on a React Flow canvas: each AI response is a node, and branching is a single click below any node or a right-click on selected text. The panel on the right edge handles model switching and response style without leaving the canvas. Sessions are portable: the app saves state as compressed .bixroute files that can be shared with others or imported to restore a previous tree.

The standout mechanic is text-selection branching. Highlighting a phrase inside any response and right-clicking lets you quote or define that phrase in a new child branch — so a dense response about, say, quantum mechanics can fracture into four parallel sub-conversations without losing the original thread. The vendor’s own examples include a metaphysics exploration tree and a choose-your-adventure visual novel for children, which signals the intended range: structured intellectual inquiry on one end, narrative exploration on the other.

Where it fits: visual thinkers, educators building branching dialogues, researchers who want to document which questions they asked and which they skipped. Where it breaks: any workflow that needs the AI to do something after the conversation — write a file, call an API, trigger a downstream step. There is no agentic layer, no tool-use loop, and no API for external integration. The canvas is a thinking surface, not a pipeline.