Code node

The Code node executes custom code within your workflow. It appears in the Advanced palette category with the description: “Execute custom code.” The node runs in a sandboxed environment with access to workflow variables and platform SDK helpers.

When to use

Transform data between nodes (reshape, filter, calculate)
Implement custom logic not covered by other nodes
Call platform SDK methods (database queries, LLM calls, DMS operations)
Parse or validate complex data structures

Configuration

Click the Code node to open its full-screen editor.

Control	Type	Description
Language	Combobox (`Python` / `JavaScript`)	Code language for this node
Variables	Button	Opens the variable browser to insert `{{variable.path}}` references
Ask AI	Button	AI-assisted code generation
Test	Button	Execute the code against test data
Exit fullscreen	Button	Return to canvas view

Code editor

Full-featured code editor with:

Line numbers and syntax highlighting
{{variable.path}} interpolation (resolved at runtime)
Autocomplete (Ctrl+Space)
Copy button

Pre-imported modules

Python: json, math, re, datetime, collections, base64, io

SDK helpers (both languages): db, llm, dms

Quick Insert snippets

Pre-built code templates accessible via buttons below the editor:

Category	Snippets
Database	SELECT Query, SELECT One Row, INSERT, UPDATE, DELETE
LLM	Chat, Multi-turn Chat, Streaming Chat, Model from Config, Embeddings
DMS	List Documents, List Folder, Download, Read Document, Get Metadata
Workflow	List Inputs, List Outputs, List Artifacts

Output settings

Setting	Type	Description
Output Variable Name	Text field (default: `result`)	Name used in `{{output.node_id.this_name}}` to access the result

Output format

Assign your result to the result variable:


result = {"status": "success", "data": processed_data, "count": 42}

Downstream nodes access it as:

{{output.<node_id>.result}} → full result object
{{output.<node_id>.result.status}} → "success"
{{output.<node_id>.result.count}} → 42

Use print() for debugging — output appears in the Test Results panel.

Available system variables

The Variables panel (left sidebar) exposes system context:

Variable	Type	Description
`runId`	string	Current execution run ID
`timestamp`	string	Execution start time
`env`	string	Environment (preview/production)
`agentId`	string	Current agent ID
`agentVersionId`	string	Current agent version
`subscriptionId`	string	Tenant ID
`currentNodeId`	string	This node’s ID
`user.id`	string	Current user ID
`user.name`	string	User display name
`user.email`	string	User email
`user.roles`	array	User’s assigned roles

Edge significance

Linear flow: single input → single output edge.

ℹ️

Code runs in a sandboxed environment (indicated by the “Sandboxed” badge on the canvas node). It cannot access the filesystem, network, or external processes directly — use the SDK helpers (db, llm, dms) for platform operations.

Common patterns

Data transformation — Reshape query results before passing to an LLM prompt.
Validation — Check data integrity and return errors if invalid.
Aggregation — Combine results from multiple upstream nodes.
SDK operations — Use db.query(), llm.chat(), or dms.list() for platform calls within custom logic.