n8n PowerPoint: generate editable .pptx slides and decks from any workflow

SlideForge is an n8n community node that turns workflow data into native, editable PowerPoint — real text boxes, tables and chart shapes in a .pptx, not a picture of a slide. Rows from Sheets, Postgres or an API go in; a finished deck comes back in about a second as binary data on the item, ready for the next node to email or upload. No AI Agent required, and no model anywhere in the render path.

Install the node

On self-hosted n8n: Settings → Community nodes → Install, enter the npm package name, accept the community-node prompt:

n8n-nodes-slideforge

Search SlideForge in the node panel and it is there:

Then add the credential: slideforge.dev/console/keys → create a key (it starts with sf_live_) → in n8n, new credential of type SlideForge API → paste. New accounts start with 60 free slides, no card.

What the node can do

The create paths also take theme, language, writing direction (including right-to-left for Arabic and Hebrew), styling, imagery and a minimum font size — set once on the deck and inherited by every slide.

Sheet to deck in three nodes

The shape almost every reporting workflow takes: Google Sheets → Code (map rows to slides) → SlideForge: Deck → Create → SlideForge: Job → Download PowerPoint → Gmail. The Code node builds a slides array:

const rows = $input.all().map(i => i.json);

return [{ json: { slides: [
  {
    form: 'hero_statement',
    headline: 'Weekly metrics report',
    context: 'Generated from the metrics sheet by n8n',
  },
  {
    form: 'kpi_metrics',
    headline: 'Where the business stands',
    source_note: 'Source: metrics sheet',
    data: {
      metrics: rows.slice(0, 5).map(r => ({
        label: r.Metric, value: r.Value, sub: r.Change,
      })),
    },
  },
] } }];

Paste that array into the node's Slides field (or wire it with an expression) and pick your options:

Not sure which layout to use? Add a Catalog → Get Many Layouts step once and read the menu, or use Slide → Create From Brief and let SlideForge route it. The layout list in Create From Layout loads live from your account, so it is never out of date.

The second node takes the job_id and hands the workflow the actual file — named, typed, and ready to attach:

Note the left-hand pane: every slide reports its form, status and fidelity. verbatim means every value you supplied is on the slide exactly as supplied. Branch on that in an IF node and your pipeline will not email a deck it cannot vouch for.

Node, MCP, or REST API

Three ways to reach the same engine from n8n, at the same price:

• The community node — typed fields, credential handling and the file as binary data. Best for deterministic reporting.
• The MCP Client node — point it at https://api.slideforge.dev/mcp/ and an AI Agent can browse the catalog and build the deck itself. Works as a standalone step too, no agent needed.
• The REST API — an HTTP Request node against /v1/render/intent/deck with a bearer key. Same contract, more plumbing. See the API docs.

PDF and image conversion to editable slides, and translation into 35 languages, ride the same account through the API and MCP surfaces.

Why a workflow wants a deterministic renderer

You can ask a language model to write python-pptx code inside n8n and it will produce a file. For a one-off that is fine. For something that runs every Monday it is the wrong shape:

• Numbers: in a generated-code path your figures travel through the model on the way to the slide. Here they are bound verbatim, and the response says so field by field.
• Time: generated code takes minutes per deck; a typed intent renders in under a second, and deck slides render in parallel.
• Stability: the same input renders the same slide, so last week's deck and this week's differ only where the data differs.
• Failure: a slide that does not fit comes back as a warning, not as silently clipped text nobody notices until the meeting.

The honest converse: if you need a bespoke layout no catalog covers, generated code is the right escape hatch — and SlideForge exposes that too, as a sandboxed python-pptx mode.

Frequently asked questions