AI Workflow Builder

How To Use

The builder provides four Drag-Drop items. Each item is described below with tips.

Add text-file attachments (plain text). When attached the file name appears on the Source item. Source items are inputs to AI Agents only.

Accepts: .txt files (plain text)
Usage: Attach supporting documents — they will be read by AI Agents when the workflow runs.

Configure an LLM, API key, and a prompt. Agents can read from Sources, use Tools, and send results to Outputs or other Agents.

Inputs: Sources, Tools, other AI Agents
Behavior: Read sources → call LLM → call Tools → forward results to output/agent items

Anything reachable via REST. Configure a URL, HTTP Options, a placeholder to inject the agent’s input, and a max number of repetitions.

URL — the tool endpoint.
Options — JSON object with method, headers, and body.
Placeholder — an attribute name where its value is the AI Agent’s LLM response, this place holder will be injected in body attribute of Options.
Repetitions — max retries until the tool accepts the input. The tool supposed to process AI Agents inputs, if the processing succeeded the tool supposed to respond with a successful HTTP response, otherwise the platform will ask the AI Agent to fix its results and then re-request the tool again.
Update Input — the tool updates the input data and returns the updated data in the response body.

Displays the final output from AI Agents. Use this to collect the processed result after agents and tools run.

While a workflow runs, events are disabled until it finishes.
To save workflows, click Save to download a JSON file, and later Load to restore it.

This platform is a serverless browser-based platform that stores no cookies.
This platform does not save or send data except the ones sent to LLMs or to Tools deliberately by the user.
This platform accesses the user storage only for saving/loading scenarios when the user calls Save/Load actions.

A comprehensive video tutorial to demonstrate the agentic AI workflow builder in action.