FlowRunner
PricingContact
Theme
Start Free

PDF Munk

Documents

Convert HTML, CSS, and live web pages into pixel-accurate PDFs and images with PDF Munk, plus a full set of PDF utilities. Agents produce finished documents without maintaining a headless browser.

11 actions API key available
PDF Munk website ↗ Platform Documentation ↗ Capability data verified 2026-08-01
The nightly billing run finishes and a set of per-account PDFs lands in file storage
Convert HTML to PDF renders each account's cover page, filling {{token}} placeholders from the account record
Merge PDFs combines the cover with that account's pages in the order the URLs are supplied
Agent compares each merged file's page count and account reference against the source manifest
Compress PDF brings the result under the mail gateway attachment limit at a strength that keeps it print-sharp
The merge manifest posts to the operations channel with page counts by account
Operations lead releases the batch, or holds any account whose page count moved since last period

What This Integration Enables

PDF Munk makes a claim most document APIs will not make: it verifies what it hands back. Every generated document is checked against its real file signature before it is stored, so a failed render cannot be saved as a valid-looking PDF and discovered later by a customer. That is a small engineering detail with a large operational consequence, because the most expensive document failure is not the one that errors loudly. It is the one that succeeds into an inbox. Templated markup is supported through {{token}} placeholders filled from a dynamic parameters object at render time, so an invoice, a shipping label and a certificate are the same action with different data.

Around that rendering core sits a full utility set: merge two to fifteen documents into one, split by page selector or into single pages or into equal chunks, compress at four strengths, lock and unlock with passwords, stamp a diagonal watermark, and parse a document back into JSON. Parse PDF is worth calling out on its own, because it reports the page count, the detected language and whether the document is a scan, and it returns plain text, layout blocks with normalized bounding boxes, tables, or a full combined pass. That makes it a credible front end for retrieval and document-understanding pipelines rather than a text dump. For a team practicing Orchestration as a Service, the value is that the whole document lifecycle sits inside the flow where the audit trail already is, instead of spread across a rendering vendor, a merge vendor and a compression vendor with three sets of logs.

Without FlowRunner

A browser fleet nobody wanted to own Document rendering means running headless Chromium somewhere, patching it, and paying for the idle capacity
Failed renders look like documents An error page or a truncated response gets saved with a .pdf extension and is only discovered by the recipient
Assembly is a separate toolchain Rendering, merging, splitting, compressing and watermarking are four vendors and four sets of credentials

With FlowRunner

Rendering is a call, not a fleet Pixel-accurate PDFs and images come back from an API with no browser to maintain
Only real documents are stored Returned bytes are checked against the file signature before anything is written to file storage
One connector covers the whole document lifecycle Render, merge, split, compress, lock, watermark and parse all run on the same credential inside one flow

Use Case Scenarios

Consolidated Monthly Statements

Each account's activity has already been rendered into per-period pages. The agent calls Convert HTML to PDF once per account to build a cover page from a single template, passing the account name, period and balance as dynamic parameters into {{token}} placeholders. Merge PDFs then assembles the cover and the period pages in the order supplied, drawing directly on the FlowRunner file URLs from the earlier steps because they are publicly downloadable. Compress PDF trims the result to fit the mail gateway limit at a strength chosen to keep tables legible. The statement that goes out is one file rather than an email with six attachments and a note apologizing for it.

Social and Preview Imagery From the Same Templates

Marketing needs an Open Graph image for every published piece, and nobody wants to open a design tool for each one. The agent calls Render HTML as Image with the same templated markup used for the page itself, applying optional CSS and Google Fonts and filling {{token}} placeholders with the title, author and date. For pages already live, Capture URL as Image screenshots the rendered result with a navigation ready state and an extra delay for asynchronous content, either at a fixed width and height or across the whole scrollable page. The images land in file storage and attach to the CMS record. The design system stays in HTML and CSS where it is already versioned.

Preparing Protected Documents for Reading

An incoming vendor statement arrives password protected, which means no downstream step can read it. Unlock PDF removes the protection using the supplied password so Parse PDF can run, returning tables and layout blocks as JSON along with the page count, the detected language and a flag for whether the document is a scan. The agent reconciles the line items against the purchase order, then Lock PDF re-applies protection before the file is filed. The moment a protected document is unlocked into shared file storage is the moment a person should have been told, so the flow tells them.

Human-in-Loop Highlight

Merge PDFs takes between two and fifteen documents and combines them in the order the URLs are supplied. The signature check that makes PDF Munk trustworthy runs on the output and confirms it is a real PDF. It cannot confirm it is the right PDF. A merge that stitched account 4471's pages behind account 4470's cover produces a perfectly valid document, passes verification, compresses cleanly, and sends another customer's balance to the wrong person. So on any batch where the merge inputs are assembled dynamically, the agent posts the manifest to the operations lead in Slack before release: "Statement batch ready. [N] merged files. [M] accounts where the page count changed against last period: [list]. Each merge combined [cover] plus [pages] in this order. Release the batch, or hold the flagged accounts?" This is exactly the kind of automation exception worth surfacing, because the failure is invisible to every automated check that exists and obvious to the person who knows the accounts.

Agent processes routinely
Detects exception requiring judgment
Clear match Continues automatically
Ambiguous Routes to human via Slack
Human decides
Agent resumes with decision

Agent Capabilities

11 actions

Document Rendering

2
  • Convert HTML to PDF Renders raw HTML markup into a PDF and stores it in FlowRunner file storage. Optional CSS is applied on top of the markup, and placeholder tokens such as {{invoice_id}} are replaced with the values supplied in Dynamic Parameters, which makes this the right action for invoices, receipts, certificates and other templated documents. Supports paper size presets, landscape orientation, background graphics, CSS @page sizing and a custom rendering viewport. The returned bytes are verified to be a real PDF before they are stored.
  • Convert URL to PDF Loads a publicly reachable web page in a headless browser and stores the rendered PDF in FlowRunner file storage. The page is fetched server side, so it must be available on the public internet without a login. Supports paper size presets, landscape orientation, background graphics, CSS @page sizing, a custom rendering viewport and a delay that lets charts, web fonts and JavaScript-loaded data settle before capture. Use it to archive a dashboard or a published report on a schedule.

Image Rendering

2
  • Render HTML as Image Renders raw HTML markup into an image and stores it in FlowRunner file storage. Optional CSS and Google Fonts are applied before rendering, and {{token}} placeholders are filled from Dynamic Parameters. Use it to turn templated HTML into social cards, notification graphics, badges and certificates without maintaining a browser.
  • Capture URL as Image Screenshots a publicly reachable web page and stores the resulting image in FlowRunner file storage. Ideal for Open Graph and social sharing images, report thumbnails and visual regression snapshots. Capture size is controlled by Width and Height, or the whole scrollable page can be captured with Full Page. A navigation ready state and an extra delay handle pages that render asynchronously.

Assembly and Optimization

3
  • Merge PDFs Combines between 2 and 15 PDF documents into a single file in the order the URLs are supplied, then stores the result in FlowRunner file storage. Every source URL must be publicly downloadable, so a FlowRunner file URL produced by an earlier step works directly. Used for assembling a cover page with generated pages, consolidating monthly invoices into one statement and building archive documents.
  • Split PDF Splits a PDF into several documents and stores every produced part in FlowRunner file storage. The split can be driven three ways: a page selector such as 1-3,5,7-10 to extract specific pages and ranges, a switch that emits one file per page, or a chunk count that divides the document into equal parts. Use it to pull out a signature page or break a batch document into per-customer statements.
  • Compress PDF Runs an optimization pass that reduces the file size of a PDF and stores the smaller document in FlowRunner file storage. Four compression strengths trade quality against size, so a lighter setting suits documents that must stay print-sharp and a stronger one suits email attachments and size-limited uploads. Compression is best effort, and already-optimized documents may shrink very little.

Protection and Marking

3
  • Lock PDF Applies password protection to a PDF and stores the encrypted document in FlowRunner file storage. Anyone opening the result must supply the password, which makes this the step to run before emailing reports, invoices or any document holding sensitive data. If the source document is already encrypted, its existing password must be supplied so the file can be opened first.
  • Unlock PDF Removes password protection from an encrypted PDF and stores the unlocked document in FlowRunner file storage. Run it before parsing, merging or splitting a protected file, since those operations cannot open an encrypted document. The correct password for the source file must be supplied.
  • Add Watermark Stamps a diagonal text watermark across a PDF or an image and stores the marked file in FlowRunner file storage. Opacity, rotation angle and font size control how prominent the stamp is, so the same action covers a faint background mark and a bold DRAFT or CONFIDENTIAL banner. The stored file keeps the format of the source document.

Reading

1
  • Parse PDF Extracts machine-readable content from a PDF and returns it as JSON rather than a file. Four modes are available: plain text for search indexing, layout blocks with normalized bounding boxes, table extraction, and a full pass combining text, layout, tables and image metadata. The response also reports page count, detected language and whether the document is a scan, which makes it a strong front end for retrieval and document-understanding flows. Extraction can be limited to a single page or a page range.

Frequently Asked Questions

What can FlowRunner do with PDF Munk?

FlowRunner agents can run Convert URL to PDF, Convert HTML to PDF, and Capture URL as Image in PDF Munk, plus 8 more actions.

Does connecting PDF Munk to FlowRunner require OAuth?

No. PDF Munk connects to FlowRunner with an API key, no OAuth flow required.

Can PDF Munk trigger a FlowRunner workflow automatically?

PDF Munk doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.

Start building with PDF Munk

$100 in credits. No card required. Connect in minutes.