FAQ

Prisment is the verification layer between human intent and AI-generated software. AI writes code faster than teams can validate architectural intent, so Prisment gives people a place to read what the AI drafted — requirements, designs, architecture, diagrams and API specs — and decide before anything is built. Open your project folder, review and edit the documents, record your decisions, then hand the agreed files back to your AI tool or team. It is free and local-first: documents are processed on your device, with no account.

Yes. You can plug in an entire project folder, and it is never sent to a server. In the desktop app, and in Chrome, Edge and other Chromium browsers, Open Local Folder mounts the folder: Prisment reads it where it sits on your computer, follows changes (for example while an AI agent writes files) and saves your edits straight back to the files. In Safari, Firefox and other browsers, use Open Folder in the Files sidebar to copy the folder’s supported files, with their subfolders, into a browser workspace kept on your device; that copy does not follow the folder on disk. Files of a type Prisment does not support are skipped. More: prisment.io/features/#feature-area-local-work

Yes. Open the project folder, choose Review beside Read and Edit, name the review and comment on passages. Prisment saves the review as files in a hidden .review folder inside the project folder, so Git carries it: commit and push, and your colleagues pull, open the folder in Prisment and reply. A synced folder such as OneDrive, Google Drive or Dropbox works the same way, and Download saves one Markdown file to send instead. Everyone who can open the folder may comment; only the person who started the review decides and approves. Prisment has no accounts or cloud storage for reviews, and an AI tool can read the review file. More: prisment.io/blog/review-documents-with-your-team-through-git/

Yes. Source Control in the activity bar uses the git installed on your computer. In the desktop app it stages, commits, amends, pushes, pulls (fast-forward only), fetches, syncs, branches, stashes, tags and clones, and your own sign-in, SSH keys and hooks apply; it sends nothing except what you press. In a browser it is read-only: the branch, the last commit and which documents changed. Rebase and force-push are not offered. More: prisment.io/features/#feature-area-review

Yes. Every feature on prisment.io — the editor, the viewers, presentation styles, local folder sync and HTML export — is free to use right now, with no account, no sign-up and no trial period, and the desktop app for macOS, Windows and Linux is a free download too. We are not charging for Prisment today; if that ever changes we will say so here first.

No. Prisment is completely client-side. All parsing and conversion are executed locally inside your web browser. Your private documents are never uploaded to any server.

Prisment is 100% client-side and serverless — your Markdown documents and user files are processed locally inside your web browser and are NEVER uploaded or stored on any server. On our hosted website (prisment.io), we use standard Google Analytics to measure aggregate page traffic and which features are used — for example that a desktop download or an HTML export happened, never the document itself — and to measure conversions for our own advertising campaigns. The desktop app collects only anonymous, aggregate usage data (a daily active signal, OS, app version, and country) to help us improve it, and you can turn it off anytime from the Help menu. The desktop app also has an optional Local AI assistant, off by default: when you turn it on, it sends document text only to a model server running on your own computer, over a loopback connection that cannot reach the internet, and it never sends that text to us or to anyone else. It can also read a web page you allow in a dialog the app shows — a link you type into a question, or, with the assistant's Follow links in documents switch on (off by default), a page your documents link to — once, for the chat or for the workspace; that request carries only the address you allowed, with no cookies and no document text beside it. The desktop app's Source Control runs the git program already on your computer when you press its buttons: Pull, Fetch, Clone and Push then talk only to the repository you chose, using your own sign-in — what you push goes where you sent it, never through us. Its Clone box can also search GitHub, GitLab and Codeberg for public repositories as you type: after a short pause, only the words you typed are sent to those three sites, with no cookies, no account, no file names and no document text, and you can turn that off in the box. In the version from our website, if the GitHub CLI (gh) is installed and signed in on your computer, Clone can also list your own repositories, private ones included, through it: Prisment runs gh, which asks GitHub with your own sign-in — Prisment never sees or stores that sign-in, nothing you type is sent for it, and without gh Clone works as before. On every platform, your document text, file names, and files are never collected or sent to analytics. Model tuning preferences stay on this computer; generation settings are sent only to that same local model server.

You can instantly drag, drop, or copy-paste Markdown (.md), MDX (.mdx), plain text (.txt), HTML (.html/.htm), CSV & TSV (.csv/.tsv), JSON (.json), YAML (.yml/.yaml), Mermaid diagram (.mmd/.mermaid), Word (.docx), PDF (.pdf), Excel (.xlsx/.xlsm) and image (.png/.jpg/.gif/.webp/.svg) files. Markdown is fully editable; a plain text file (.txt) opens and edits exactly as it is, with no rendering at all — what you type is what gets saved, so indentation, alignment and any stray Markdown characters survive untouched; JSON and YAML edit as syntax-coloured source; a Mermaid file edits as source beside its live pan-and-zoom diagram; CSV opens as a sortable grid and HTML as a view-only preview. YAML OpenAPI/Swagger specs render as browsable API docs (any other YAML shows as code). Word documents (.docx) open as a reading view — headings, lists, tables, links, images and footnotes, without page layout, fonts or colours — and convert to Markdown beside the original. PDFs (.pdf) open page by page in a built-in viewer and convert to Markdown as well: a tagged PDF keeps its headings, lists and tables; an untagged one is rebuilt from the text layout, with page headers and footers removed; a scanned PDF has no text to convert. Pictures in a Word or PDF document are left out for the leanest Markdown, or kept as image files in a folder beside it and shown in the preview — your choice, remembered. Excel workbooks (.xlsx/.xlsm) open as an interactive multi-sheet grid and convert each worksheet to CSV beside the original. A picture converts to PNG, JPG or WebP — resized or compressed if you like — beside the original, copies to the clipboard as an image or as a Markdown link, and zooms; an SVG is changed by the local assistant — its labels, colours and boxes — rather than in an XML editor, and the assistant can export it for you too.

Yes. Click a picture and press Size and alignment: align it left, centre or right, choose Fit or 25, 50, 75 or 100 % of the column, move the slider or drag the corner; Reset puts it back. With no choice, a picture fits the page by its shape: never wider than the text, never taller than most of a screen, never enlarged. Prisment writes the choice after the picture, as {width=50% align=center}. GitHub and CommonMark have no size syntax, so another tool shows that as text; an HTML image tag with a width is the portable form, and Prisment reads it the same way. More: prisment.io/blog/resize-and-align-images-in-markdown/

Yes. Prisment converts Word (.docx) and PDF (.pdf) documents directly to clean Markdown right inside your browser with 100% client-side execution. Headings, lists, tables, and formatting are preserved and saved beside your original file, ready for AI tools or documentation workflows.

Yes. Excel workbooks (.xlsx and .xlsm) can be converted into standard RFC 4180 CSV tables client-side. You can preview all sheets in an interactive data grid and convert the active sheet or all sheets to CSV without sending financial or sensitive data to external servers.

Excel workbooks (.xlsx and .xlsm) open as a reading grid — one tab per sheet, hidden sheets marked — not as a spreadsheet editor, so support is deliberately partial. What comes through: cell values exactly as Excel displays them, so currency, percentages and dates keep their number format; the author's own highlights and cell fills, bold, italic, underline, font colours and any explicit left/centre/right alignment; and merged cells, showing their value once. A long sheet loads its first 5,000 rows with a “Load all rows” control, and a workbook over 50 MB is not opened. What is not shown: conditional formatting (colour scales, data bars, icon sets) and table styles, because Excel computes those when it opens the file rather than storing them per cell; fonts, font sizes and cell borders, so the grid keeps following your chosen presentation style; and charts, pivot tables, embedded images and comments. Formulas display their last saved result — one saved without a result shows blank — and macros in an .xlsm are never read or run. For complete fidelity, use “Open in default app” to hand the original file to Excel or Numbers.

AI tools and large language models (LLMs) read Markdown and CSV at a fraction of the token cost of binary Word, PDF, or Excel files. Converting them locally leaves lean, readable files right beside the originals, stripping bloated XML and binary overhead while keeping all underlying information intact.

Yes, in the desktop app only. It answers questions about the folder you opened, and edits documents when you ask, using a model that runs on your own computer through Ollama or LM Studio. We recommend Qwen2.5-Coder 14B; a 3B model such as Llama 3.2 3B runs on less memory but can leave a section out of a summary or guess where the folder has no answer. Prisment finds the passages that answer and sends them to that model over a loopback connection that cannot reach another machine, so nothing is uploaded and no account is needed. A web page is read only after you allow it, and a question is not answered without the page it links. How it works, and what to expect from each model: prisment.io/blog/local-ai-assistant-for-your-documents/

Yes! You can open, edit, and format MDX (.mdx) files in Prisment. Custom MDX React component rendering is not yet supported, but full custom component rendering is currently on our product roadmap.

Yes! You can drag and drop multiple Markdown files into a single multi-document workspace to manage, view, edit, and switch between files instantly.

Mostly. An exported page is a single HTML file that fetches its theme and runtime from cdn.prisment.io the first time it is opened, then caches them — so it opens anywhere with a connection, and offline from then on. On a machine that has never been online it falls back to showing your document as plain, copyable text instead of a blank page. For a copy that is guaranteed to work with no network, use Print / PDF.

You can copy the raw parsed HTML code to your clipboard, save the document as a print-optimized PDF, or download a fully responsive, standalone HTML web page.