Skip to content
DogubakoDogubakoDogubako

Introduction

Dogubako (道具箱, “toolbox”) replaces the dozens of single-purpose utility websites and small apps you reach for every week: the image converter, the PDF merger, the video compressor, the JSON formatter, the subnet calculator. Open it, find the tool, drop in a file, and you are done.

It is free and open-source software under the GNU AGPL v3.0 or later. There are no accounts for desktop use, no telemetry and no paywalls.

What it can do

Section titled “What it can do”

Dogubako ships 249 tools in twelve categories:

Category Tools Highlights
Images 30 Convert (HEIC, AVIF, WebP…), batch resize and compress, crop, watermark, metadata, compare, icons and favicons, face blur with manual correction, redaction, OCR, background removal, upscaling
Video 27 Convert and compress (including to a target size and with hardware encoders), trim, cut, merge, crop, rotate, GIF ⇄ video, stabilize, frames, subtitles
Audio 21 Convert, EBU R128 loudness normalization, trim, split, merge, silence removal, waveforms, spectrograms, tags, denoise, EQ
Speech & Voice 9 Offline transcription with whisper.cpp to TXT, SRT, VTT, JSON or Markdown, language detection, speaker turns, text to speech with Piper
PDF & Documents 27 Merge, split, reorder, rotate, compress, OCR to searchable PDF, extract text and images, watermark, page numbers, encrypt, forms, Office and Markdown conversion
Files 23 Hashes and checksum checks, archives with bomb and traversal protection, duplicates, large files, folder compare, batch rename with preview
Text 27 Case, counts, sort and dedupe, diff, regex replace, encodings, escaping, Markdown ⇄ HTML, text extraction from Office, EPUB and PDF
Developer 34 JSON, YAML, TOML and XML format and convert, Base64, URL and hex, UUIDs, passphrases, hashes and HMAC, timestamps, regex tester, JSONPath and jq, JWT, SQL format, cron, QR and barcodes
Data 12 CSV view, filter, sort, stats, dedupe and clean, spreadsheets, JSON ⇄ CSV, schema validation, export to XLSX, SQL or Markdown
Network & Web 18 DNS, RDAP and WHOIS, HTTP header and security audit, TLS certificates, URL tools, CIDR and subnet planning, connectivity
Design 9 Palettes and palette extraction, WCAG contrast, gradients, canvas resize, QR styling
System 12 Overview dashboard, CPU, memory, disks, battery health, hardware, processes, temperatures

The tool catalog lists every tool with its options, the engines it needs and the platforms it runs on.

Five ways to use it

Section titled “Five ways to use it”

All five share one engine and one catalog, so a tool behaves the same everywhere.

  1. Desktop app, released for macOS on Apple silicon (it builds on Windows and Linux too). Everything runs in-process on your machine. There is no server and no network port.
  2. Self-hosted web app. Run one server with all the engines installed and use it from any browser on your network, including phones.
  3. Command line. Every tool is a dogubako command, for scripts and quick one-off jobs.
  4. HTTP API, with an OpenAPI 3.1 document generated from the live catalog.
  5. MCP server for AI agents, either locally over stdio or shared over HTTP with a scoped token per agent.

What makes it different

Section titled “What makes it different”
  • Built on mature open-source engines. FFmpeg, ImageMagick, ExifTool, qpdf, Poppler, Ghostscript, Tesseract, whisper.cpp, Piper, LibreOffice, Pandoc and 7-Zip do the heavy lifting. About half the catalog needs none of them and runs as fast native Rust code. The macOS app and the Docker image bring the engines with them.
  • Plain words, if you want them. Type “make these photos smaller” and a model you choose runs the right tools, asking before anything that touches the network or your folders. Off by default. See Plain words.
  • One implementation, every interface. Each tool is a declarative spec and one function. Forms, CLI flags, API schemas and MCP tool definitions are generated from the spec. See Architecture.
  • Agent-ready, not agent-first. A shared server can give an image bot permission to resize and blur, and a media bot permission to transcode and transcribe, without either touching the other’s files or your system tools.
  • Secure by construction. No shell is ever invoked. External programs run with timeouts, resource limits and optional sandboxing. Files are addressed by random ids and checked for ownership. Network tools refuse private addresses. Everything is audit-logged. See Security.

Next steps

Section titled “Next steps”