JustFill MCP Server

Let AI agents (Claude, ChatGPT, n8n — any MCP client) detect, review and fill PDF form fields through justfill.app.

Fill PDF forms in ChatGPT

Install the published JustFill PDF Forms plugin from the ChatGPT plugin directory, then connect your JustFill account. This hosted path does not require installing the Python client or configuring an MCP server manually.

Try this prompt with the one-page, six-field sample:

Use JustFill to open https://justfill.app/examples/chatgpt/project-intake.pdf.
List its fields and show a preview. Do not fill anything yet.

Ask for a filled preview with fictional values and approve it before generating the PDF. Save the reviewed layout, then open the same original blank PDF for the next client or project. Your saved template is available in the same JustFill account; generated files follow that account's plan.

Try the two-project example and download the real editable results. The example includes prompts for two fictional clients, the blank PDF, and two generated PDFs with their free-output watermark.

Hosted server — no local installation

If your MCP client supports remote servers with OAuth, use this server URL:

https://justfill.app/api/mcp

Connect from your client, sign in to JustFill, and review the access request. You do not need to install the Python package or copy an API key for this path. Access can be revoked in JustFill under Account → API Keys.

Connection guide. The client is MIT-licensed; PDF processing uses your JustFill account's allowance. For a first check, use a blank form and fictional values, and review the filled preview before exporting. Save the reviewed layout to reuse it with new data when the same form comes up again.

Excel or CSV batch workflow

If the source data is already in a spreadsheet and you need one filled copy of the same existing PDF per row, an MCP client is optional. The guided browser workflow imports XLSX or CSV, maps columns to reviewed PDF fields, previews each record, and exports the approved PDFs in a ZIP.

Try the five-row PDF mail merge sample — no card or sales call.

Import-ready n8n workflows

Start with the deterministic workflow JSON in this repository. It collects a PDF and JSON payload, reuses the reviewed field names saved for that exact form, fills the original layout and returns a temporary download link.

The workflow is listed in the n8n template library. For new runs, import the current strict-mapping JSON from this repository, which includes the synthetic fixture and validation evidence alongside it. The workflow is free to download; running it uses a JustFill account and its PDF-processing allowance. It uses built-in nodes, not the separate JustFill community node.

The repository also includes the exact deterministic workflow JSON, synthetic test PDF, production evidence and a separate two-pass vision workflow for an unfamiliar form. Both call the hosted MCP endpoint with standard HTTP Request nodes and can be inspected before adding credentials.

Inspect the source workflows and evidence, or follow the step-by-step n8n setup.

Business example: recurring supplier intake

An operations team can keep the supplier's required intake PDF unchanged, save its reviewed field layout once, and let n8n map approved vendor data from a webhook or CRM record into that exact form. The workflow returns a temporary filled-PDF link that can be reviewed before it is uploaded to Drive, attached to a draft email, or written back to the vendor record. The repository's synthetic supplier-intake PDF exercises this exact path without customer data.

Gemini CLI extension

Install the same reviewed MCP tools plus the included PDF workflow guidance:

gemini extensions install https://github.com/mrmaciej1/justfill-mcp

The extension manifest lives at the repository root and uses the published justfill-mcp package. Gemini CLI asks for normal third-party extension consent before enabling it.

Why agents can trust it

ML confidence is calibrated: the detector's raw scores are not probabilities (its server-side filter accepts boxes from raw ~0.02 and auto-accepts at raw 0.15), so they are mapped onto 0–1 to mean what you'd expect — ≥0.75 "detector is sure", 0.4–0.75 "probably right, glance at the preview", <0.4 "borderline accept, verify". The raw detector score is kept on each field as raw_score.

The correction loop (render_preview → add/update/remove_field) exists precisely because ML detection has false positives and negatives. A false positive costs nothing (leave it unfilled or remove it); a false negative is visible on the preview and fixable with one add_field call. Once reviewed, save_template makes every future fill of that form deterministic.

Local client setup

uv tool install justfill-mcp

Authorize once (opens the browser, one click while logged in to justfill.app):

justfill-mcp login

Then the config needs no credentials at all:

{
  "mcpServers": {
    "justfill": { "command": "justfill-mcp" }
  }
}

For a zero-install configuration, use uvx directly:

{
  "mcpServers": {
    "justfill": {
      "command": "uvx",
      "args": ["justfill-mcp"]
    }
  }
}

Alternatives, in the order the server checks them:

1. JUSTFILL_API_KEY env — create a key at justfill.app → Account → API Keys and put "env": {"JUSTFILL_API_KEY": "jf_live_…"} in the config.
2. The key saved by justfill-mcp login (~/.config/justfill/credentials.json).
3. JUSTFILL_EMAIL + JUSTFILL_PASSWORD — legacy fallback; an API key is better (no password in config files, revocable per client, never expires mid-session).

Tools

• open_pdf(path, min_confidence=0.0, max_pages=10, force_detect=False) — template → AcroForm → ML resolution order. Accepts scanned images too (jpg/png/tiff → converted to PDF, deterministically, so templates still match). force_detect=True ignores a saved template and re-runs ML.
• render_preview(page_index) — page image with labeled field boxes (blue = deterministic, green/orange/red = ML confidence)
• render_filled_preview(values, page_index) — the same page with your values drawn in place (checkboxes get an X). Costs no fills — check before you fill.
• list_fields(page_index?)
• add_field(x, y, w, h, name, page_index, field_type, align?, vertical_align?) — coords in % of page, top-left origin
• update_field(field_id, …) / remove_field(field_id)
• update_fields([{field_id, …}, …]) / remove_fields([ids]) — batch versions
• prune_fields(field_type?, confidence_below?, width_below?, height_below?, page_index?, exclude_ids?) — bulk-delete detection noise in one call (criteria AND-ed, removed ids returned)
• fill_pdf(values, output_path, flatten=True) — values = {field_id: text}; responds with warnings for values that will be shrunk/truncated to fit
• save_template(name) — persist the reviewed layout for deterministic repeat fills
• list_templates()

Text alignment: align = left|center|right, vertical_align = top|middle|bottom — set per field (e.g. right for RTL forms, center for boxed digits). Persisted in templates.

Example agent flow

open_pdf("~/forms/w-9.pdf")            → acroform, 27 fields, confidence 1.0
fill_pdf({"f1": "Jane Doe", …}, "~/out/w-9-filled.pdf")

open_pdf("~/forms/scan.jpg")           → converted to PDF; ml, 34 fields
render_preview(0)                      → agent sees noise + one missed line
prune_fields(field_type="cell", width_below=3)   → 16 removed in one call
add_field(x=18, y=62.5, w=40, h=3, name="Phone")
render_filled_preview({…})             → values sit right, no overflow
fill_pdf({…}, "~/out/filled.pdf")
save_template("Client intake form")    → next time: deterministic

Notes

• Auth is a regular justfill.app account; tokens auto-refresh on expiry.
• Usage and document-output rules are enforced by the same account service as the web app. fill_pdf reports whether the output is clean or watermarked.
• One PDF open at a time per server session (by design — keeps ids stable).
• This repository mirrors released versions of the MCP client (development happens in a private monorepo alongside the justfill.app backend). Bug reports and feature requests are very welcome in the issue tracker here.