Skip to main content
Glama
formfeed

formfeed

Official

Server Configuration

Describes the environment variables required to run the server.

NameRequiredDescriptionDefault
FORMFEED_REGIONNoAPI region of the workspace.eu
FORMFEED_API_KEYYesFormfeed API key from the app (API keys). ff_test_… renders free with a watermark, ff_live_… consumes units.

Instructions

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

Formfeed renders PDFs and images from HTML templates, and fills Word and PowerPoint templates (as DOCX, PPTX or PDF). Start with list_templates, read the data shape with get_template_schema, check data with validate_template, then call render and hand the download_url to the user. An image render also returns the picture itself, so look at it before handing it on. For an e-invoice (ZUGFeRD / Factur-X), put the invoice into data._invoice, check it with check_invoice, and call render with einvoice set; the template prints the same block. validate_einvoice says which rules an e-invoice XML passed, by which validator, also one Formfeed did not make. markdown_to_pdf turns a Markdown document, such as a report you wrote, into a PDF without a template. convert_to_pdf turns office documents into PDFs. For a PDF that already exists, a render or one you are given: get_pdf_info reads its pages and form fields, fill_pdf_form fills the form by those names, redact_pdf removes text for good and checks that it is gone, split_pdf and compress_pdf cut and shrink it, and pdf_to_image draws a page and hands the picture back so you can look at it. To change a template or write a new one, read it with get_template_source and save it with save_template: it stays a draft until a person publishes it in Formfeed. get_workspace says which workspace this server works in and what the period has left.

Capabilities

Features and capabilities supported by this server

Protocol revision2025-11-25

CapabilityDetails
tools
{
  "listChanged": true
}

Tools

Functions exposed to the LLM to take actions

NameDescription
list_templatesA

Lists the templates of the workspace this server works in (the one chosen at sign-in, or the API key's), most recently changed first: id, slug, name, kind (pdf, image, docx for Word, pptx for PowerPoint), engine, tags and published_version. Call it first to find the slug that get_template_schema, validate_template and render take; skip it when the user already named the template. Returns every match in one answer, without archived templates, and an empty list when nothing matches. A template whose published_version is null has never been published, so render cannot use it yet. Read-only and free.

get_template_schemaA

Returns the JSON Schema of the data object a template expects and the sample data it was designed with, from its published version, else its latest draft. Call it after list_templates and before render or validate_template, to build data that fits; ad-hoc html has no stored schema to read. When the template stores no schema, one is inferred from the sample data: treat that as a guide, not a contract. Read-only and free; an unknown slug returns an error.

get_template_sourceA

Returns what a template version is made of: the body markup (html), css, head, settings (paper, margins, header, footer, fonts), sample data, data sets, data schema, translations and partials, with the version's checksum. Call it before save_template to change a template the user named, and save what you edited with that checksum as base_checksum. Reads the latest version by default, the newest draft a change builds on; published reads what renders use. Word and PowerPoint templates are documents: for them it returns the settings and data, not the document. Read-only and free.

validate_templateA

Checks a template and its data without rendering: free, no units, nothing stored, no document produced. Call it before render when you assembled the data yourself, or after writing ad-hoc html. With template, the API checks the published version (else the latest, or the version you name) as a render would: syntax, header and footer, errors only this data triggers, and the data against the stored JSON Schema (a wrong type or a missing required field is an error). With html and engine it compiles offline and reports syntax errors, unknown filters and variables missing from data. Returns ok and diagnostics (severity, code, message, plus path, line and column, or part and paragraph for Word and PowerPoint). Warnings leave ok true: settle each before rendering.

save_templateA

Saves a template as a draft, never published: renders keep using the published version until a person publishes the draft in Formfeed. With template and base_checksum it saves a new version of that template: read it with get_template_source first, change what the user asked for, and pass only the parts you changed (the others are kept from the version you read). With new_template and html it creates a template whose version 1 is the draft. change_note says what changed; the person reads it before publishing. Returns the saved version, the check of the draft with its sample data (fix every error and save again) and review_url, the page where the person reviews and publishes it: hand that to the user. Needs an API key with template:write, or, for a signed-in agent, the person's permission under Account → Connected agents in Formfeed. Word and PowerPoint templates cannot be saved here. Free; a new template counts towards the plan's template limit.

check_invoiceA

Checks the _invoice block of an e-invoice offline, as the API does before it renders one: the fields EN 16931 asks for, the totals to the cent, VAT categories and code lists, and France's rules for an invoice between two French businesses. Free, no units, nothing sent. Returns ok and problems, each with the field, the rule (BR-…) where one applies and a message. Nothing is computed for you: change the values a problem names, check again, then call render with einvoice and the block as data._invoice. The validator that runs during the render holds the file to the full rule set and to PDF/A-3b as well.

validate_einvoiceA

Says which rules an e-invoice passed, by which validator and release: the file is held to the rule set its specification identifier (BT-24) names, Factur-X and ZUGFeRD by Mustang, XRechnung (CII and UBL) by the KoSIT validator, and each check (syntax, schema, every named rule set) comes back with its own result and findings. Pass exactly one source: render_id, an e-invoice this workspace rendered, which is answered from the report the render keeps and costs nothing; path, an XML file on this machine; or xml_base64, the XML itself. An XML takes one of the organisation's validations of the day (from the Starter plan; Free validates no uploads). Up to 4 MB of CII or UBL; a PDF is not taken. result is passed, failed or not_checked: not_checked means no rule set here is for the file (reason and explanation say why, such as an identifier it does not know), not that the file is wrong. The report says which rules a file passed; it does not say whether the document is a proper invoice for VAT purposes. To check an _invoice block before rendering it, use check_invoice.

renderA

Creates a document by filling a stored template, or ad-hoc html, with data: a PDF or image, or DOCX, PPTX or PDF from Word and PowerPoint templates. Returns the render with status and download_url; for an image (png, jpg or webp) up to 2 MB the result also carries the picture itself, so look at what you made before handing it on (a PDF comes as its link). Use it to make a document from data; to turn an existing office file into a PDF, use convert_to_pdf, and for a Markdown document, markdown_to_pdf. Stored templates render their published version, so one whose published_version is null fails with template_not_found. Every call is a new render: a live key or a signed-in agent consumes units, a test key is free (watermarked on the Free plan). A render that fails returns status failed and an error with code and message; validate_template finds most causes beforehand. With einvoice the PDF becomes a ZUGFeRD / Factur-X e-invoice built from data._invoice (check the block with check_invoice first): the render then carries einvoice, the validator's report, and a refused invoice names the fields and rules to fix. The link expires at expires_at; get_render issues a fresh one.

markdown_to_pdfA

Turns a Markdown document into a PDF without a template, e.g. a report, notes or an answer you wrote, and returns the render with status and download_url. It reads GitHub Flavored Markdown (tables, task lists, strikethrough) and footnotes; a YAML front matter block gives the title, author, subject and keywords, else the first heading is the title; a line \newpage starts a new page. Raw HTML in the text is shown as text. A picture named without an address (Logo) is a file of the workspace library, uploaded in Formfeed; with path, a picture beside the file is sent along instead. The headings become the PDF bookmarks by default, and pdf_ua checks the file against PDF/UA-1 before it is delivered. To fill a template with data, use render; to convert an office file, use convert_to_pdf. Every call is a new render: one unit per started five pages on a live key or for a signed-in agent, free with a test key; every plan.

convert_to_pdfA

Converts an existing Word, Excel, PowerPoint, OpenDocument, RTF or HTML document (up to 20 MB) to a PDF as it is, without filling in data, and returns the render with status and download_url. To make a document from a template and data, use render instead. Pass exactly one source: render_id, the render of a Word or PowerPoint template (output docx or pptx); path, a file on this machine; or file_base64 with file_name. Every call is a new render that counts units like a PDF; the Free plan cannot convert (Starter and above). The document is converted and not kept.

get_pdf_infoA

Reads a PDF without changing it: the page count and the size of every page, whether it is password-protected, its document properties, and its form, with every field by its full name, label, type, value and the options it takes. Read it before fill_pdf_form, which takes the fields by these names, and before redact_pdf or split_pdf, which work on these pages. Pass exactly one PDF: source, a render or library file of the workspace; path, a file on this machine; or file_base64. Free on every plan; it creates no render, and an uploaded PDF is read and not kept.

split_pdfA

Cuts a PDF into parts, every N pages each or one part per entry of ranges, and returns the render with download_url: a ZIP of the parts, named after the source (contract-1.pdf, contract-2.pdf and so on), or the PDF itself when there is one part. get_pdf_info gives the page count first. Pass exactly one PDF: source, a render or library file of the workspace; path, a file on this machine; or file_base64. Starter plan and above; 0.5 units on a live key. A PDF/A or a password-protected PDF is refused.

compress_pdfA

Makes a PDF smaller and returns the render with download_url; a PDF that cannot be made smaller comes back as it was. For a preview picture of a page instead, use pdf_to_image. Pass exactly one PDF: source, a render or library file of the workspace; path, a file on this machine; or file_base64. Starter plan and above; 0.5 units on a live key. A PDF/A or a password-protected PDF is refused.

pdf_to_imageA

Draws pages of a PDF as PNG or JPEG pictures and returns the render with download_url: one page gives the picture, which this tool also hands back so you can look at it (up to 2 MB), several pages a ZIP of pictures named by page. At most 100 pages and 120 million pixels a call, about 13 A4 pages at 300 dpi. Use it to see a page before redact_pdf, or for thumbnails. Pass exactly one PDF: source, a render or library file of the workspace; path, a file on this machine; or file_base64. Starter plan and above; one unit per picture on a live key.

redact_pdfA

Removes what lies under rectangles of the pages, and every place where a text stands, from the file itself and not just from view: text, the pixels of images under a rectangle, and links and comments it touches. The result is read again: a text that can still be found fails the call with redaction_incomplete, which names rectangles and texts by their number, and nothing is delivered. Rectangles are points from the top-left corner of the page as a viewer shows it, 72 to an inch; get_pdf_info gives the page sizes and pdf_to_image a picture of the page. A text is found whatever its spacing, ignoring case unless match_case is true. A black box is drawn over each removed place, another colour with fill, none with fill null. Pass exactly one PDF: source, a render or library file of the workspace; path, a file on this machine; or file_base64. Starter plan and above; 0.5 units on a live key. The texts are never stored.

fill_pdf_formA

Fills the fields of a PDF form by their full names, as get_pdf_info lists them, and returns the render with download_url and form: how many fields were filled, which were left out, and which hold a calculation or display format that no viewer ran. The form stays fillable, or flatten draws the values into the pages. A name the form lacks or a value its field does not take fails the call with form_field_unknown or form_value_invalid, naming the field, unless strict is false. Pass exactly one PDF: source, a render or library file of the workspace; path, a file on this machine; or file_base64. Starter plan and above; 0.5 units on a live key. The values are never stored; a signed form and an XFA form are refused.

get_renderA

Returns the current state of a render: status (queued, rendering, succeeded or failed), download_url with expires_at, page_count, units and error, and the picture of a finished image render up to 2 MB. Use it to follow a render started with wait false, or to get a fresh link once the old one has expired; render, convert_to_pdf and the PDF tools already return this for the render they just made. Read-only and free; an unknown id returns an error.

get_workspaceA

Returns the workspace and organisation this server acts in, the plan, the environment (live or test) and the units of the current period: included, used and the balance of prepaid units. Call it when the user asks where a document will land or how many renders are left, or when list_templates returns nothing you expected. A signed-in agent works in the workspace chosen when it was connected, which the person changes in Formfeed under Account → Connected agents; an API key works in the workspace it was made for. Read-only and free.

Prompts

Interactive templates invoked by user choice

NameDescription

No prompts

Resources

Contextual data attached and managed by the client

NameDescription

No resources

TDQS

A4.4/5.0

Scored across 18 tools

Disambiguation4/5

The boundaries are mostly explicit: render, markdown_to_pdf and convert_to_pdf each state when to use the other, and the three validators (validate_template, check_invoice, validate_einvoice) target clearly different artifacts (template+data, an _invoice block, an XML file). Residual confusion is limited to the PDF-producing family and the validator family, where an agent must read the fine print rather than glance at the name. No two tools are genuinely interchangeable.

Naming Consistency4/5

Everything is snake_case and almost all names are verb_noun (list_templates, split_pdf, redact_pdf, validate_einvoice). Minor deviations: 'render' is a bare verb with no object, and the PDF tools mix suffix style (split_pdf, compress_pdf) with prefix style (pdf_to_image, get_pdf_info). Readable and predictable overall.

Tool Count4/5

18 tools is on the heavy side but each maps to a real capability of a document platform: rendering, seven PDF utilities, template read/write/validate, e-invoicing checks, and workspace inspection. Nothing looks like a filler alias or a duplicate endpoint. It sits just below the point where the surface would need grouping.

Completeness4/5

Coverage is strong: template discovery, schema, source, draft save, validation, render lifecycle (render/get_render), and a solid PDF toolbox. Obvious gaps are merge (split_pdf exists but has no inverse) and text extraction, and template deletion/publishing is deliberately left to the web UI. Those are workable around but would cause the occasional dead end.

Maintenance

ActivityActive
ResponsivenessNo recent community issues