JustFill PDF Forms

Fill any PDF form with AI agents — ML field detection on scans, an agent-review loop, reusable templates, and native AcroForm fill via justfill.app.

CommunityOnlineNo sign-inGlobalFreeRead-only

What it can do

  • Request File Upload: Get a one-time upload slot to hand this server a PDF (or form photo). Use this whenever the document lives on the user's device or was attached to this chat — tool arguments canno
  • Open Pdf: Open a PDF (or an image of a form: JPG/PNG/TIFF) and detect its fields. Provide the document ONE of three ways: - `url`: a public link to the document (best when it exists online), - `upload
  • List Fields: List the working fields of a workspace (optionally one page only).

What data it sees

Do you need an account

No: the server works without sign-in

Fill any PDF form with AI agents — ML field detection on scans, an agent-review loop, reusable templates, and native AcroForm fill via justfill.app. Works with Claude, ChatGPT, n8n and any MCP client. Hosted OAuth endpoint, no install.

Server tool list (14)

Raw names from tools/list. Only developers need these.

request_file_uploadGet a one-time upload slot to hand this server a PDF (or form photo). Use this whenever the document lives on the user's device or was attached to this chat — tool arguments cannot carry file bytes reliably. NORMAL PATH (chat assistants): give the user the returned upload_url, ask them to open it and drop the file there (one click), then call open_pdf with upload_token. Your sandbox almost never has network access to justfill.app, so don't burn a turn trying to POST it yourself, and never fall back to base64 or shrink the file — that breaks detection. AUTOMATION ONLY: if you are a headless runtime with open network egress (n8n, a server shell), you may POST the file to post_url instead (`curl -sf -F "file=@document.pdf" <post_url>`) and skip the user step, then open_pdf with upload_token. The slot is single-use and expires in 30 minutes.
open_pdfOpen a PDF (or an image of a form: JPG/PNG/TIFF) and detect its fields. Provide the document ONE of three ways: - `url`: a public link to the document (best when it exists online), - `upload_token`: from request_file_upload — USE THIS for files the user attached to the chat or has on their device; do NOT try to inline big base64 blobs through the model, they get truncated, - `file_base64`: only for small files (< ~200 KB) you already hold as exact bytes. NEVER compress, strip fonts from, or rebuild a larger document to squeeze it under that limit — a modified file breaks field detection and no longer matches its template; use upload_token instead. Resolution order: saved template (deterministic) -> embedded AcroForm -> ML detection with per-field confidence. force_detect=True skips templates. Returns a workspace_id — pass it to every other tool.
list_fieldsList the working fields of a workspace (optionally one page only).
render_previewRender a page with the field boxes drawn on it — verify placement. Blue = deterministic (template/AcroForm/agent), green/orange/red = ML confidence bands. Labels show field ids.
render_filled_previewPreview how the filled page will look BEFORE fill_pdf. Costs nothing. `values` maps field id -> text (checkboxes tick on "yes"/"x"/"true"). Typography is approximate; use it for placement/overflow checks.
add_fieldAdd a field the detector missed. Coordinates in % of page, top-left origin.
update_fieldMove/resize/rename a field or set its text alignment.
update_fieldsBatch update_field. Each item: {"field_id": ..., x?, y?, w?, h?, name?, field_type?, align?, vertical_align?}. Unknown ids are reported back.
remove_fieldDelete a field that isn't a real input (a false positive).
remove_fieldsBatch remove_field.
prune_fieldsBulk-delete fields matching ALL given criteria (detection noise). Thresholds are strictly-less-than; w/h in % of page. Returns removed ids.
fill_pdfFill the PDF and return a download link (valid until the workspace expires). `values` maps field id -> text (checkboxes: "yes"/"x"/"true" to tick). Read `warnings` (values that will be shrunk/truncated) and `output_mode`: "watermarked" means the account's free fills are used up — TELL THE USER before delivering; a paid plan at justfill.app/billing removes it.
save_templatePersist the reviewed field layout — future open_pdf of this exact file returns these fields with confidence 1.0 and no ML pass. Before saving, give every field a short semantic name (update_fields with name=..., e.g. 'age_score', 'total_abcd2') — names are stored in the template, so the next session maps values by meaning instead of guessing from coordinates. Also remove false positives first (remove_fields).
list_templatesList saved templates on this account (name + field count + hash).