A visual flow builder that compiles to an execution graph
Seven archetypes across seven channels, 130 node types, and a canvas that shows execution order rather than a decorative diagram. What you publish is what runs.
A production graph, mid-call
Pulses follow execution order through a live flow: entry, recognition fork, tool call, queue decision, disposition.
Live canvas preview — pulses follow the order nodes actually execute in.
How flows get built
Archetype first
The archetype decides which nodes exist at all — inbound support, outbound campaign, appointment, collections, verification, survey, routing only. The channel then removes the rest.
AI co-pilot draft
Describe the flow in a paragraph and the co-pilot lays out a first graph: nodes, edges, prompts and tool stubs, ready to edit.
Dry-run self-healing
Before publish, the flow runs against sample conversations. Dangling edges, unmapped variables and unreachable nodes are reported and, where unambiguous, repaired.
What a node actually does
Nodes are typed: media nodes control the call leg, inference nodes open a recognition or generation turn, integration nodes call your systems, and control nodes branch on variables the flow has collected.
Because the graph is the unit of execution, it is also the unit of metering and audit. An invoice line maps to the nodes that ran, and a conversation record replays in node order with inputs and outputs preserved.
- Versioned flows with rollback to any published revision
- Per-node timeout, retry and fallback edge
- Variables scoped to the conversation, typed at declaration
- Node-level execution log on every conversation record
Draft your first flow with the co-pilot
Bring a call script and we will have a working graph on the canvas during the session.