iiinova — where your digital self keeps evolving

PDFops

@pdfops.dev

64 followers 310 following 692 posts

Hosted PDF API for AI agents, n8n and Python. Fill AcroForm fields, merge PDFs, inspect forms, generate invoices. Deterministic output, no headless browser. MCP server + SDKs, free key at pdfops.dev

Posts Replies Media

PDFops @pdfops.dev · 01/10/2026

New on the blog: Merging a Letter invoice with an A4 attachment in pypdf, and the page size nobody normalizes. pypdf's append() copies each input page's own /MediaBox into the merged output as-is, so a Letter invoice merged with an A4 attachment comes out with two page sizes. Why →

Merging a Letter invoice with an A4 attachment in pypdf, and the page size nobody normalizes — PDFops

pypdf's append() copies each input page's own /MediaBox into the merged output as-is, so a Letter invoice merged with an A4 attachment comes

PDFops @pdfops.dev · 30/09/2026

The fill endpoint takes multipart/form-data, not a JSON body. In Python that's the PDF in files= and the field map in data= as json.dumps(fields), a JSON string, not a dict. Get that shape right and the rest is one requests.post. Working snippet →

Fill a PDF in Python | PDFops

Deterministic AcroForm fill in one HTTP call, no native dependencies.

PDFops @pdfops.dev · 30/09/2026

Agents re-run tool calls: timeouts, re-plans, the same step twice. If the fill tool is deterministic, three identical pdf_fill calls give you three byte-identical files, so you dedupe by hash instead of guessing which copy is the real one. PDF tools over MCP →

PDF tools for AI agents over MCP | PDFops

Give an agent pdf_inspect, pdf_fill and pdf_merge, and why it should inspect first.

PDFops @pdfops.dev · 29/09/2026

Byte-identical output means you can name a PDF after its own hash. Use sha256(bytes) as the R2 or S3 key: identical fills collapse into one object, and "did we already generate this one?" becomes a HEAD request. It only works if your generator is byte-stable →

The case for deterministic PDF filling | PDFops

Why same input to byte-identical output changes how you store, cache and test PDFs.

PDFops @pdfops.dev · 29/09/2026

Contract templates change when legal edits them. If the field names stay the same, your calling code doesn't have to. Version the template file and put that version in a footer field at issue time, so every PDF you ever sent says which template made it. The pattern →

Contract PDFs from onboarding flows | PDFops

Fill a contract template from form data, then hand the PDF to your e-sign tool.

PDFops @pdfops.dev · 28/09/2026

Filling a hybrid AcroForm/XFA PDF still returns 200 and a PDF, so the status code can't tell you the XFA layer was dropped. We set an x-pdfops-warning: xfa-layer-dropped response header instead. Read it in your handler: older XFA-first Adobe viewers show the blank template →

Fill a PDF in Node.js | PDFops

Deterministic AcroForm fill from your backend in one HTTP call.

PDFops @pdfops.dev · 28/09/2026

If your PDF needs conditionals, loops and partials, a real templating engine like PDFMonkey's Liquid is the right tool and PDFops isn't. We fill fields on a fixed-layout PDF that already exists, nothing more. Pick by whether the layout changes per document →

PDFops vs PDFMonkey | PDFops

When an HTML templating engine fits, and when filling an existing PDF does.

000

PDFops @pdfops.dev · 27/09/2026

Filling an invoice template means the layout is fixed: a template with 10 line-item fields holds 10 lines, and line 11 has nowhere to go. Size it for your real worst case, or build a base PDF per invoice first. That's the price of never re-rendering the layout. Both patterns →

Invoice PDFs at scale | PDFops

Fill an invoice template from a webhook and get identical bytes every time.

PDFops @pdfops.dev · 27/09/2026

PDFShift bills overage per credit past your plan. PDFops does the opposite today: hit your quota and you get a 429 until the 1st, so the bill is always the sticker price. The tradeoff is real, your code has to handle that 429. Which failure mode you'd rather have decides it →

PDFops vs PDFShift | PDFops

Hosted PDF API comparison: form fill and merge versus HTML rendering.

000

PDFops @pdfops.dev · 26/09/2026

Quota math: a fill is one request and a merge is another, so a filled-then-merged bundle costs two. On the $16 plan (4,000 requests) that's 2,000 bundles a month. Only 2xx responses count, so failed calls while you debug are free. Count requests per document →

PDFops | PDF fill and merge API

Deterministic PDF fill and merge over HTTP, from $16/mo or free to try.

PDFops @pdfops.dev · 26/09/2026

A field name that isn't in the PDF gets a 400 unknown_field with the name in the message, not a 200 and a blank box. A value longer than the field's maxLength gets exceeds_max_length instead of being truncated. Typos and overflows fail loudly, so you can catch them in CI →

PDFops playground

Fill and merge PDFs in your browser, no signup.

000

PDFops @pdfops.dev · 25/09/2026

Field map gotcha: every value has to be a string, numbers and checkbox states included. {"qty": 3} gets a 400 invalid_field_value that names the field. {"qty": "3"} fills. Convert once where you build the map, not field by field at call time →

Fill a PDF in JavaScript | PDFops

Fill AcroForm fields from the browser or a backend with one HTTP call.

PDFops @pdfops.dev · 25/09/2026

The Inspector reads your PDF's AcroForm fields entirely in the browser tab. Nothing is uploaded, so a real customer's form never leaves your machine while you work out field names. It also writes the fill call for you, curl, fetch or Worker, with those names already in it →

PDF Form-Field Inspector | PDFops

List a PDF's AcroForm fields in your browser and copy the exact fill call.

PDFops @pdfops.dev · 24/09/2026

New on the blog: Giving an AI agent PDF tools over MCP, and why it has to inspect before it fills. An MCP server hands an agent pdf_inspect, pdf_fill, and pdf_merge tools directly, no client code to write. The field-name problem that makes inspecting first non-optional, and what →

Giving an AI agent PDF tools over MCP, and why it has to inspect before it fills — PDFops

An MCP server hands an agent pdf_inspect, pdf_fill, and pdf_merge tools directly, no client code to write. The field-name problem that makes

PDFops @pdfops.dev · 24/09/2026

Fill a PDF's fields after it's already been digitally signed and most viewers will flag the signature as invalid — any edit to a signed file, even one field, falls outside the byte range the signature covers. Fill first, sign last, never the other way around →

Signed contracts from a Tally webhook

Fill a contract template from Tally answers and email back a signed PDF, on one Worker.

PDFops @pdfops.dev · 23/09/2026

Not every 'pick one' group in a PDF is a radio group. Some forms fake it with independent checkboxes laid out to look exclusive — same visual style, different field names — so setting one true doesn't clear the others. Check the type before you assume selecting is exclusive →

Fill a PDF in JavaScript

Build a correct field map from real AcroForm data, in the browser or on a server.

000

PDFops @pdfops.dev · 23/09/2026

Marking a field Required in Acrobat only makes Acrobat itself nag before a human hits save. Neither the PDF format nor /fill-form enforces it — you can submit a field map that skips a required field entirely and the call still succeeds. Validate that yourself first →

Fill a PDF in Python

Native libs vs one POST with a JSON field map, compared with working code.

PDFops @pdfops.dev · 22/09/2026

A text field with the Password flag set only tells Acrobat to mask what you type on screen — the stored /V value in the file is a plain string, no hashing, no encryption. Treat it like any other visible field when you decide what belongs in a field map →

PDF field Inspector

Paste a PDF, see every AcroForm field, its type, flags, and current value.

PDFops @pdfops.dev · 22/09/2026

Text fields, checkboxes, dropdowns, radio groups, option lists — /fill-form fills all five. Push buttons and signature fields aren't silently skipped, they 400 with unsupported_field_type, so a typo'd field name and an unfillable field type don't look like the same failure →

AcroForm field Inspector — PDFops

See every AcroForm field's real name, type, and widget count before you fill.

PDFops @pdfops.dev · 21/09/2026

Merging five PDFs into one drops each source's outline tree — the merged file is a fresh document with only page objects copied in, so whatever table-of-contents bookmarks they had don't survive. Build a new one against the merged page count if you need it →

PDFops playground

Fill and merge in the browser and see exactly what gets sent.

PDFops @pdfops.dev · 21/09/2026

Merge concatenates page objects, not visual orientation — each page keeps whatever /Rotate value its source PDF gave it. Feed it a landscape scan and a portrait template and the merged file comes back with both, exactly as mixed as the inputs were →

Reports — PDFops

Turn a recurring data payload into a client-ready PDF report with one API call.

000

PDFops @pdfops.dev · 20/09/2026

Merge doesn't try to preserve N source files' timestamps — the output gets a fixed epoch CreationDate/ModDate and a single Producer string, every time. However many PDFs go in, whenever they were made, the merged file's own metadata is always identical →

Merging PDFs in a Bun server

Page order, quota checks, and the 429 you'll eventually hit — in Bun.serve().

000

PDFops @pdfops.dev · 20/09/2026

Loading a PDF for /fill-form uses updateMetadata: false, so the file's own CreationDate and ModDate pass through untouched instead of pdf-lib's default of stamping wall-clock time on save. One line of config, one less thing breaking byte-identical output →

The case for deterministic PDF filling

Same input, byte-identical output — hash it, cache it, sign it.

PDFops @pdfops.dev · 19/09/2026

flatten=true on /fill-form bakes field values into page content and drops the AcroForm — that's the whole guarantee. It is not tamper evidence, not a signature, not compliance. Flatten to stop edits; sign separately if you need proof nothing changed after →

Contracts — PDFops

Fill a contract template's fields deterministically, then route it to e-sign.

000

PDFops @pdfops.dev · 19/09/2026

Every /fill-form call regenerates every field's appearance stream with an embedded Helvetica, unconditionally — so a programmatically-built AcroForm whose DA points at a font it never embedded still renders filled, not blank. No NeedAppearances flag to remember →

Fill a PDF in Node.js — PDFops

One POST, a JSON field map, deterministic bytes back — appearances included.

PDFops @pdfops.dev · 18/09/2026

PDFops' /fill-form takes a checkbox value as the plain string "true" or "false" — never the field's actual on-state export code like /Yes or /1. pdf-lib resolves that mapping internally, so you never inspect a checkbox just to check it. Try it live →

PDFops playground

Try a fill or merge call, then copy the exact snippet for your stack.

PDFops @pdfops.dev · 18/09/2026

New on the blog: Filling a checkbox with pypdf, and the on-state string it never validates. pypdf's update_page_form_field_values() sets a checkbox with zero validation, so a wrong on-state string leaves it silently unchecked. The state it actually needs, and the flag that hides →

Filling a checkbox with pypdf, and the on-state string it never validates — PDFops

pypdf's update_page_form_field_values() sets a checkbox with zero validation, so a wrong on-state string leaves it silently unchecked. The s

PDFops @pdfops.dev · 17/09/2026

One field name can have multiple widget annotations — an initials box repeated on every page of a 12-page lease is one field, not twelve. Set its value once and every visible instance updates together. The Inspector lists it as a single entry, not a duplicate per page →

AcroForm field Inspector — PDFops

See every AcroForm field's real name, type, and widget count before you fill.

000

PDFops @pdfops.dev · 17/09/2026

DocRaptor's async mode hands back a job ID and expects you to poll or wait on a callback, because rendering HTML/CSS at scale can be slow. Filling fields on a PDF that already exists has no render pass to wait on, so the call is just request in, bytes out, no job to track →

PDFops vs DocRaptor

Synchronous fill and merge, compared against DocRaptor's async render jobs.

000

PDFops @pdfops.dev · 16/09/2026

Same fill call, four snippets: the playground shows curl, Node, Python, and a Cloudflare Worker fetch side by side for whatever you just ran. Copy whichever matches your stack and the request is already correct — no translating a curl example into your language by hand →

PDFops playground

Try a fill or merge call, then copy the exact snippet for your stack.

000

PDFops @pdfops.dev · 16/09/2026

An invoice PDF that's about to be emailed doesn't need to touch disk in between: fill-form returns bytes, and most transactional email APIs take an attachment as base64 or a buffer directly. Generate and send in the same request handler, nothing written, nothing to clean up →

Invoice PDFs — PDFops

Fill an invoice template and attach the bytes to an email, no intermediate file.

PDFops @pdfops.dev · 15/09/2026

New on the blog: Getting started with PDFops. Sign up, grab a key, and make your first fill-form and merge call — the walkthrough covers both primitives end to end, with the exact request bodies, before you build anything on top of them →

Getting started — PDFops

First fill-form and merge calls, request bodies included, start to finish.

PDFops @pdfops.dev · 15/09/2026

The response is just bytes — an ArrayBuffer you can pipe straight into an S3 PutObject or an R2 bucket without ever writing a temp file to disk. No cleanup step, no /tmp quota to hit on a serverless box with a few hundred MB to work with →

Fill a PDF in Node — PDFops

Bytes back, no temp file — pipe the response straight to object storage.

000

PDFops @pdfops.dev · 14/09/2026

The field map in Python is just a dict — no schema class, no dataclass, no generated types to keep in sync with the PDF. Paste the form into the Inspector, copy the field names it lists, build the dict, POST it. The PDF's own AcroForm tree is the only schema that matters →

Fill a PDF in Python — PDFops

A plain dict as the field map — no native deps, no generated types.

000

PDFops @pdfops.dev · 14/09/2026

Anvil's plans are seat- and volume-tiered, priced for a team building forms inside their editor. If you already have the PDF and just need fields filled from a script, that tier structure is overhead for a job that's one HTTP call. What the two pricing models actually assume →

PDFops vs Anvil

Pay-per-call fill and merge, compared against Anvil's editor-first, seat-priced model.

PDFops @pdfops.dev · 13/09/2026

A PDF signature field (/FT /Sig) holds a cryptographic signature dictionary, not a text string — there's no value to POST into it. Fill-form only touches text, checkbox, and choice fields; a signature field just passes through empty, waiting for whatever signs it →

Contract PDFs — PDFops

Fill a contract template's fields deterministically, then route it to e-sign.

000

PDFops @pdfops.dev · 13/09/2026

The Inspector lists fields in the PDF's internal object order, not left-to-right, top-to-bottom reading order. Field #1 in the list can be a footer widget on page 4 while the form's first visible box is field #38. Don't assume position from the index →

AcroForm field Inspector — PDFops

See every AcroForm field's real name, type, and value — in the order the file stores them.

PDFops @pdfops.dev · 12/09/2026

PDF/A, the archival format compliance teams ask for, bans embedded JavaScript outright. If a report template has a calculated field — a script summing line items — that script gets stripped on save, so the total has to be computed before you fill, not left to run in the viewer →

Report PDFs — PDFops

Fill recurring report templates deterministically, compliance-safe by default.

000

PDFops @pdfops.dev · 12/09/2026

Netlify Edge Functions run on Deno under the hood, so a native PDF binary or a headless browser is a non-starter there. A fetch call with a JSON field map doesn't care what's under the hood — same code path as Node or a Worker →

Fill a PDF in JavaScript — PDFops

One fetch call fills PDF form fields from any JS runtime, edge or not.

000

PDFops @pdfops.dev · 11/09/2026

New on the blog: Packing-slip PDFs from a fulfillment webhook, and the redelivery that duplicates them. A fulfillment webhook's order-shipped event can redeliver under retry, and a naive handler fills the same packing slip twice. The atomic-claim pattern, on Vercel with Upstash R →

Packing-slip PDFs from a fulfillment webhook, and the redelivery that duplicates them — PDFops

A fulfillment webhook's order-shipped event can redeliver under retry, and a naive handler fills the same packing slip twice. The atomic-cla

PDFops @pdfops.dev · 11/09/2026

Government forms love auto-generated field names like topmostSubform[0].Page1[0].f1_01[0] — unreadable on their own. Most also carry a /TU tooltip string, the human label the form author actually typed. The Inspector surfaces it next to the real name. #webdev

AcroForm field Inspector — PDFops

Paste a PDF, see every field's real type and export values before you guess at a field map.

000

PDFops @pdfops.dev · 10/09/2026

No sales call, no demo booking, no "contact us for pricing": sign up, get a key, make your first fill-form call in about a minute. That's the entire onboarding funnel for a $0.004-per-call API. →

PDFops — deterministic PDF fill & merge API

Free tier, 100 requests/month, rate-limit headers on every response.

PDFops @pdfops.dev · 10/09/2026

New on the blog: Fill and merge PDFs in your browser. Useful past the demo: a KYC packet with an applicant's ID scan never has to leave their machine for a one-off job, and the same tab shows you the exact API call to make once it stops being one-off →

Fill and merge PDFs in your browser

Free, no-signup browser tool for filling and merging PDFs — no backend required.

PDFops @pdfops.dev · 09/09/2026

New on the blog: Monthly PDF bundles on a Cron Trigger. If the trigger fires twice for the same period — a retry, a redeploy mid-run — fill and merge are plain stateless calls, so rerunning today's batch doesn't double-bill or double-mail anyone. Idempotent by default →

Monthly PDF bundles on a Cron Trigger

Per-customer statement bundles built and delivered on a schedule, no standing server.

000

PDFops @pdfops.dev · 09/09/2026

New on the blog: Auto-fill a W-9 from a webhook. Typeform sends snake_case answer keys; the W-9's AcroForm fields are named nothing like that. The actual work in this pipeline isn't the fill call, it's the field map translating one naming scheme into the other →

Auto-fill a W-9 from a webhook

Turn onboarding-form answers straight into a filled W-9, on Vercel Edge.

PDFops @pdfops.dev · 08/09/2026

Wrote up why the name is PDFops and not PDF-generator: the product treats documents like infrastructure — stateless calls, idempotent retries, byte-identical output — not like a creative rendering pipeline. That framing decided the entire API surface →

Why I built PDFops

An operator's account of why hosted PDF generation needed a rebuild for edge runtimes.

PDFops @pdfops.dev · 08/09/2026

DocRaptor is built to lay out HTML/CSS from scratch — great when you're designing the page. A W-9 or a court form isn't a layout problem, it's fields that already have positions; filling them doesn't touch the page at all. Where the two tools actually differ →

PDFops vs DocRaptor

Rendering HTML layouts vs filling the fields of a fixed-layout PDF.

PDFops @pdfops.dev · 07/09/2026

PDFShift bills for render time because it's booting a browser to lay out HTML on every request. If your input is a PDF that already exists and only the field values change, there's no layout pass to bill for — the cost model doesn't have that variable. Compared →

PDFops vs PDFShift

HTML-to-PDF rendering vs deterministic form-fill — different jobs, compared.

PDFops @pdfops.dev · 07/09/2026

Filling a PDF from Python doesn't need asyncio, threads, or worrying about the GIL — there's no CPU-bound rendering happening in your process at all. It's one POST and one response, so a sync requests call and an async client cost the same. Both shown → #webdev

Fill a PDF in Python

Native libs vs one POST with a JSON field map, compared with working code.

000

PDFops @pdfops.dev · 06/09/2026

An auditor re-running last quarter's compliance report should get the exact same PDF back, byte for byte, given the same source data — not a document that merely looks the same. That's a determinism requirement, not a formatting one. #webdev

Fill reports with PDFops

A JSON payload and a template that doesn't change — no re-render per run.

000