JSON Tools

Parse, transform, and manipulate JSON data.

Overview

The JSON Tools connector is a built-in utility for inspecting, transforming, and reshaping JSON data inside a workflow. It exposes primitives for parsing strings into objects, querying nested values with path expressions, merging documents, computing diffs, and flattening or unflattening deep structures, all without leaving the canvas.

It sits between connectors that produce JSON (HTTP, MongoDB, Shopify raw-graphql, webhook triggers) and connectors that consume it (Klaviyo, NetSuite, MySQL, Resend). Use it to pick fields out of a noisy API response, normalize payloads before a database insert, or validate that an AI agent's structured output matches the shape your downstream system expects.

What You Can Do

The JSON connector exposes these tools:

parse - Parse a JSON string into an object.
stringify - Serialize an object to a JSON string.
prettify - Format JSON with indentation for human readability.
minify - Strip whitespace from a JSON string.
validate - Check that a string is syntactically valid JSON.
query - Run a JSONPath-style query and return matching values.
get - Read a single value at a dot-path (e.g. order.customer.email).
set - Write a value at a dot-path, creating intermediate keys as needed.
pick - Return a new object with only the listed keys.
omit - Return a new object with the listed keys removed.
keys - List the keys of an object.
values - List the values of an object.
merge - Deep-merge two or more JSON objects.
diff - Compute the structural difference between two JSON documents.
flatten - Collapse a nested object into a single-level map of dot-paths to values.
unflatten - Expand a flat dot-path map back into a nested object.

Authentication and Setup

No connection or authentication is required. These tools are built into the platform and available in every workflow by default - just drop a Connector node onto the canvas and pick the tool you need.

Using in a Workflow

Add a Connector node, select JSON Tools, and pick a mode:

Direct Mode - Best for deterministic transforms. Call get with a known path, or pick with a fixed key list, and pass the result into the next node.
Agent Mode - Useful when a downstream AI step needs to reshape variable payloads on the fly (e.g. "pull just the fields I need from this webhook").

For high-volume pipelines, prefer Direct Mode - it skips the agent loop and runs as a single deterministic step.

Tips

Use get with a dot-path instead of parse + manual traversal. It's a single step and survives renames better.
Flatten before diffing when comparing nested config or settings documents - the diff is easier to inspect line-by-line.
Validate AI output with validate before passing it to another connector. AI models occasionally emit trailing commas or unescaped newlines.
Use pick early in a workflow to reduce the payload size that flows through the rest of the canvas.

Common Pitfalls

String vs object - Most other connectors hand you a JSON object already; you only need parse when reading from a raw string body (e.g. a webhook with text/plain content-type).
Dot-paths and arrays - Use bracket notation (items[0].sku) when stepping into arrays; bare dots only work for object keys.
Deep merge semantics - merge overwrites scalars and concatenates arrays. If you need a different array strategy, do it explicitly.
Flatten key collisions - Flattening an object that contains literal dots in its keys produces ambiguous paths. Sanitize keys first.

Common Use Cases

For technical API details and field specifications, see the JSON Tools documentation.