v1.2.0 — the launch release

v1.2.0 — the launch release

Four languages, a full docs site, and an independently audited AI interface.

• Speaks your language — the whole UI is now localized in Spanish, German, and French (plus English), with a one-click switcher that remembers your choice.
• Docs site — every guide, the complete CLI reference, the MCP server reference, the security model, and FAQ, in one searchable site (docs/site, built with VitePress).
• Audited for agents — an independent re-audit of the CLI and MCP server against the project's threat model found and fixed a real sandbox escape (a symlink inside --root could break containment), added a --read-only MCP mode, and hardened untrusted-text delimiters. Full report: docs/SECURITY_ASSESSMENT_AGENTS.md.
• Cleaner and clearer — stale Netlify/Vercel deploy configs removed, expanded CLI/MCP local-install docs, CI green, new social preview banner.

Try it in your browser — files never leave your device, nothing is tracked: https://navigatorslab.com/pdf-studio

Self-host in one line:

docker run -d -p 8080:80 --name pdf-studio ghcr.io/kayforkind/navigatorslab-pdf-studio:1.2.0
# → http://localhost:8080

v1.1.0 — MCP server + CLI

Agentic interfaces

• pdfstudio CLI (packages/cli): 9 commands — info, extract-text, search, edit-text, redact, merge, split, rotate, pages. Unix-friendly: - stdin, -o - stdout, --json, exit codes 0/1/2.
• MCP server (packages/mcp): 10 tools over stdio for AI agents (pdf_info, pdf_extract_text, pdf_search_text, pdf_edit_text, pdf_redact_text, pdf_redact_rect, pdf_merge, pdf_split, pdf_rotate, pdf_pages). Path containment, no shell-outs, extracted text marked untrusted against prompt injection.
• Same privacy guarantee as the app: document bytes never leave the machine.

Engine fixes (also live on the site)

• Fixed invalid PDF output: TJ-array operands now emitted with the TJ operator (was Tj — text vanished in strict parsers); deletions emit () Tj instead of a bare operator.
• One-click wipe now clears all stored keys (found in the independent security audit — repo scored A−).

See the README's new Agentic usage section for install, CLI examples, and claude mcp add config.

v1.0.0

PDF Studio v1.0.0

First stable release — the free, open-source PDF editor that runs 100% in your browser. No uploads, no accounts, no watermarks, no page limits.

What it does

• True in-place text editing: rewrites the page content stream itself, exports real vector text
• Fill real AcroForms (text, checkboxes, radios, dropdowns), then flatten them
• Sign and annotate: signature pad, highlight, pen, arrows, shapes, sticky notes, image stamps
• True redaction: covered text is deleted from the file at export, not painted over
• OCR via on-device Tesseract (EN/ES/FR/DE) — nothing uploaded, ever
• On-device AI Q&A over your document (WebLLM, weights cached after first fetch)
• Revision diff, page organize (reorder, rotate, merge, split, duplicate, delete), export stamps
• PWA: installs to phone or desktop, works offline, autosaves locally

In this release

• Mobile experience production-ready (touch drawing fixed, single-column export dialog, reachable Save, 44px touch targets)
• MIT license detection fixed on the repo page
• Docker self-host one-liner: docker run -d -p 8080:80 ghcr.io/kayforkind/navigatorslab-pdf-studio:latest
• Issue and PR templates, third-party notices

Coming in v1.1.0

• MCP server + CLI for agentic workflows

Live: https://navigatorslab.com/pdf-studio/