Every step in your workflow produces an output. What that output looks like depends on the step type.
Viewing outputs
After a run completes (or while it’s running), click on any completed step in the canvas. The side panel opens and shows the result.
Text outputs
Ask AI and other AI steps produce text — the AI’s written response.
In the side panel, you’ll see:
- The response rendered as formatted text (with headings, bullet points, etc.)
- A Completed badge with a green checkmark
This is the most common output type. It’s what you get when you ask the AI a question or give it a task.
Structured outputs
Some steps produce structured data — organized information with specific fields, not just free text. This is what makes StewAI powerful for data workflows.
When a step has a schema attached (see Schemas), the AI is required to respond with data that matches specific fields. For example, a “Company Analysis” schema might require company_name, industry, and competitors.
In the side panel, structured outputs have two viewing tabs:
- Graph — A visual tree view that lets you expand and collapse fields
- JSON — The raw data in a code block
When to use structured outputs
Use them when you need:
- Specific fields — “Give me exactly these 5 data points”
- Consistent shape — Every run produces the same fields, making downstream steps reliable
- Export-ready data — Structured data can be directly exported to spreadsheets
File outputs
Steps like Export Table and Document produce files that save automatically to the Pantry — StewAI’s file storage.
When configuring these steps, you choose:
- Format —
.xlsx, .csv, .pdf, .docx, or .txt
- Destination — a Pantry folder and filename
After the run completes, open the Pantry from the sidebar to find and download your files.
| Step | File types | What it does |
|---|
| Export Table | .xlsx (Excel), .csv | Converts structured data into a spreadsheet |
| Document | .pdf, .docx (Word), .txt | Renders text content into a formatted document |
How outputs connect
The real power of StewAI is that one step’s output becomes another step’s input. When you insert a token pill from Step A into Step B’s prompt, Step B receives whatever Step A produced.
For example:
- Input step → user types “Tesla”
- Ask AI step → receives “Tesla”, produces a text analysis
- AI Chef step → receives the text analysis, produces structured data with fields like
competitors, market_cap
- Export Table step → receives the structured data, produces an Excel file
Each step in the chain automatically waits for the previous step to finish before running.
Token pill colors tell you the output type. Orange pills are text outputs. Green pills are structured data fields. This helps you know what kind of data you’re working with.
