Expression Language

Any string field on a node can mix plain text with expressions: {{ ... }} blocks that run JavaScript against the current workflow data. An expression editor shows autocomplete for the built-in variables below and a live preview of the result, evaluated against the node's last "Execute step" run.

Hello {{ $json.name }}, you have {{ $json.count }} new orders.

If the field is a single {{ ... }} block with nothing else around it, its evaluated value keeps its real type (number, object, array) instead of being converted to a string.

Built-in variables

Variable	What it gives you
$json	The current item's data — the main thing most expressions read.
$node["Node Name"]	That node's output, for reaching back further than the immediately preceding node.
$items()	Every item produced by the current node's input, as an array.
$input	The raw input passed into the current node.
$workflow	Workflow-level info: id, name, active version.
$execution	The current run: id, mode, start time.
$env	Environment variables your workspace has explicitly allow-listed. There is no passthrough to the host's own environment.
$vars	Workflow Variables, set per workspace.
$now / $today	The current date/time, as a Luxon DateTime — see below.

Older ${...} expressions still work

FluidGrids originally shipped a simpler ${output.field}-style template (scopes: output, TriggerInput, WorkflowState, RunContext, item). Workflow versions saved with that syntax keep running exactly as before — {{ }} is a separate, additive expression language, not a replacement that breaks old versions.

Dates with Luxon

$now and $today are Luxon DateTime objects, so the usual Luxon methods are available:

{{ $now.plus({ days: 7 }).toISODate() }}
{{ $today.toFormat('yyyy-MM-dd') }}

Querying data with JMESPath

For reaching into a nested structure without writing a chain of .map()/.filter(), call $jmespath(data, query):

{{ $jmespath($json.orders, "[?status=='paid'].total") }}

See the JMESPath specification for query syntax.

Helper methods

Values carry a small set of extra, n8n-style helper methods on top of normal JavaScript:

Type	Methods
String	.isEmpty(), .isNotEmpty(), .extractEmail(), .extractDomain(), .toDateTime()
Array	.isEmpty(), .isNotEmpty(), .chunk(size), .pluck(key), .unique()
Object	.isEmpty(), .isNotEmpty()

{{ $json.email.extractDomain() }}
{{ $json.tags.isEmpty() ? 'untagged' : $json.tags.pluck('label') }}

Sandboxing

Expressions run in an isolated evaluator with no access to the host filesystem, network, or process — only the built-in variables and helper methods above are reachable.

What's not here yet

Per-item lineage ($("Node Name").item resolving to the specific upstream item a value traces back to, and a Code node "run once per item" mode) needs the executor to dispatch per item rather than per combined batch. That is tracked separately as the per-item execution model work.