Process and data blocks

Transform

Use Transform to reshape input data into a new JSON structure. It is the safest first block after a trigger when downstream steps need predictable field names. It accepts any JSON-compatible block output and returns the object or array described by the mapping. Use it to rename fields, compose payloads, select nested values and prepare strict input for Web Request, Queue Action, Issue Action or AI Agent.

Code

Use Code for custom logic that is too specific for configuration-only blocks. Prefer built-in blocks for common operations because they are easier to inspect and maintain. Code accepts the current workflow context and should return serializable JSON. Keep output explicit: { "ok": true, "items": [...] } is easier to route than a plain string or provider-specific object leaked from a library.

Web Request

Use Web Request to call external SaaS APIs or internal services. Configure URL, method, headers, query parameters and body, and keep secrets in Vault values. It accepts templated strings, JSON bodies and optional file references from Drive, OneDrive or S3 integrations. It returns status, headers and parsed body when the response is JSON. Format the body in Transform first when the API expects a strict schema.

DB Query

Use DB Query to run controlled database reads or checks through a configured database connection. It accepts a query plus parameters from workflow data and returns rows, affected counts or database-specific result metadata. Prefer parameter fields over string-built SQL when user input is involved.

Read File and Save File

Read File loads data from project storage or integrations such as Drive, OneDrive or S3. Save File writes workflow output to storage, for example /exports/{{run.date}}/{{run.id}}.json. Read File returns file metadata and content or a reference depending on file size and provider. Save File accepts text, JSON or binary references from earlier blocks and returns the storage path, provider ID and public or private access metadata.

Wait for all

Use Wait for all when several branches must finish before the workflow continues. It accepts outputs from multiple incoming edges and returns a combined object keyed by branch or connection. Add a Transform after it when the next block needs one flat payload.

AI Agent

Use AI Agent inside a workflow to summarize, classify, extract, generate or decide based on configured prompts and allowed AI tools. It accepts prompt templates, optional system prompt, model/source settings, response schema and memory settings. It returns model text, structured JSON when schema output is configured, tool-call metadata and usage information when available.