A generated file is an artifact a workflow run creates: a report, a CSV, a rendered audio clip. It starts as a value a block produces and becomes a workspace file when a File block writes it to the Files store. Once saved, it has a name, a size, and a URL, and any later run can read it back.
Like a build pipeline that produces artifacts and stores them in an artifact repository, a workflow produces file artifacts that land in your workspace Files store, indexed by ID and shared across every workflow.
What produces file content
Most generated files start as the output of an earlier block. Two patterns are common.
A block returns text content you want to keep. An Agent writes an analysis or summary; a Function builds a CSV or formats a report. That text is part of the block's output, read by reference as <agent.content> or <function.result>. To make it a file, you pass that value into a File block set to Write.
A block returns a file object directly. ElevenLabs text-to-speech returns an audioFile; an image generator returns a generated image; the File block's own Read operation returns parsed files. These are already UserFile objects (an object with id, name, url, size, and type), so they appear in the output panel as files with no Write step. You can hand one to a downstream block, or write its content to the Files store to persist it.
A block's output is remembered under the block's name for the rest of the run, and a later block reads it by key, like <elevenlabs.audioFile> or <agent.content>. Producing content and saving it are two separate steps: the producing block holds the value, the File block persists it.
Saving content with the File block
The File block's Write operation creates a new workspace file. It takes two required fields: a fileName (like report.md) and the content string to store. It returns the saved file's details.
To save an Agent's analysis, connect Agent into a File block, set the operation to Write, and reference the Agent's output: fileName = research_summary.md, content = <agent.content>. When the workflow runs, the File block writes the file and produces:
id: the canonical file ID, used to fetch the file later.name: the final file name after any deduplication (see below).size: the byte count.url: an absolute URL to download or preview the file.
The content type is detected from the file extension: .csv becomes text/csv, .pdf becomes application/pdf, .md becomes text/markdown. To set it yourself, fill in contentType in advanced mode.
If a file with the same name already exists in the workspace, Write does not overwrite it. It appends a numeric suffix instead: a second data.csv is saved as data (1).csv, a third as data (2).csv. To add to an existing file rather than create a new one, use the Append operation, which writes content to the end of a named file.
Where the file goes
A saved file lands in the workspace Files store, the same place uploads live. It shows up in the Files panel in the sidebar with its name, size, type icon, and modified date, grouped by category: document, image, audio, video, or code. From there you can preview it, rename it, move it into a folder, or download it by URL.
Files are scoped to the workspace, not to one workflow. A file written by one workflow is visible to every other workflow in the same workspace. Any later run can read it back with the File block's Read or Get operation, or by selecting it in a file picker.
During a run, the file also appears in the output panel as the File block's output, shown as a structured object with its id, name, url, and size. The output panel shows you one run as it happens, while the Files store is where the file stays afterward.
Returning a generated file from a deployment
When a workflow is deployed as an API, a generated file can be part of the response. Reference the file in a Response block, or include its ID in the object you return. The caller uses the url or id to fetch the file from the workspace store. The file itself stays in the Files store, and the response carries a pointer to it, not the bytes.