Step-by-step guide to using the 'code' node for custom JavaScript logic.

henryf · 2026-07-28T19:25:55Z

Been using Relevance AI's workflow builder for a few months. The 'code' node is the most powerful tool in the box if you need to step outside pre-built nodes. Here's how I use it for custom logic, like transforming data or calling external APIs. * **Accessing inputs**: Your node inputs are in `inputs`. Use `inputs.someInputKey`. * **Returning outputs**: Your code must return an object. Map your outputs to the keys you defined in the node config. * **Async/await**: Almost everything is async. Write your function accordingly. * **Example - Status Check**: Need to branch based on a previous node's result? Code node is perfect. Example: I had a workflow that needed to check a deployment status from an external system and route accordingly. ```javascript // This is pseudocode for the logic async function userCode(inputs) { const status = inputs.deploymentStatus; if (status === "SUCCESS") { return { route: "successPath", message: "Proceed" }; } else if (status === "FAILED") { return { route: "failurePath", message: `Failed with status: ${status}` }; } return { route: "pendingPath", message: "Hold" }; } ``` Key thing: keep it focused. Use it for logic, not for building entire applications. If your script gets long, consider making a custom tool instead.

Page 2 / 2 Prev

Relevance AI Reviews

Last Post by cost.analyst.liam 1 hour ago

16 Posts

16 Users

0 Reactions

2 Views

RSS

cost.analyst.liam

(@cost_analyst_liam)

Reputable Member

Joined: 4 months ago

Posts: 166

29/07/2026 8:09 am

The cost differential between a native filter node and a hand-rolled script filter can be staggering when you run the numbers. A native node executes compiled code, while your custom loop is interpreted in their sandbox, with compute units charged per instruction.

You've identified the core optimization: treat the platform's pre-built nodes as managed services, and your custom code as expensive Lambda functions. The pricing model often incentivizes using their higher-level abstractions. The hidden lock-in isn't just the inability to export the logic, it's the architectural incentive to design your workflows around their specific nodes to avoid cost blowback.

Always check the data transfer costs.

ReplyQuote

Page 2 / 2 Prev