Data flow and tokens
How outputs flow between nodes and how to reference upstream data with template tokens.
Workflows is in Beta. The UI and API may change. The sidebar entry is currently disabled for most workspaces — your account team enables it for pilot orgs.
Every node writes an output to the run context. Downstream nodes reference earlier outputs through template tokens or, for the Transform node, through JSONata expressions over the whole context.
The run context
The run context is a key/value map of node id to node output. The trigger node's output is the first entry — a parsed body for webhooks, the matched event payload for Composio, or an empty object for Schedule triggers. The context grows as nodes complete. A node never sees a sibling's output until that sibling has finished; execution order is determined by the graph edges.
Template tokens
Most node fields accept tokens of the form {{ node_id.path.to.value }}. The TokenInput in the Properties drawer offers autocomplete from any upstream node's output schema. Paths that are empty or missing at run time render as empty strings rather than causing a run failure.
SchemaParamsForm
For nodes whose action has a published input schema — integration actions in particular — the Properties drawer renders a form generated from that schema. Object fields render as collapsible nested groups. Validation is non-blocking: required fields show inline warnings, but the form saves regardless. Unfilled required fields cause a run-time failure at the node.
Transform (JSONata) — when tokens aren't enough
The Transform node evaluates a single JSONata expression with the full run context as input. Use it to reshape objects, filter and aggregate arrays, or join outputs from multiple upstream nodes. Reference any node's output by its id: node_id.field. Expression errors surface as run-time failures with the JSONata cause message preserved in the run detail.