Use Privaku tools from an agent.

Discovery

/tools/llms.txt is a plain-text summary. /tools/agents.json lists every tool with JSON Schema inputs and outputs.

JavaScript API in the page

// on any /tools/ page, after load
privakuTools.tools   // [..., "photo-privacy-check"]
// run(tool, fileOrFiles, options) or run(tool, { files, ...options }); returns JSON plus Blob(s), no UI needed
await privakuTools.inspect(file)   // { format, groups: { gps:{present,lat,lon}, exif:{make,model,date,...}, xmp, iptc, icc, comment, c2pa } }
const [r] = await privakuTools.run("remove-photo-location", file)                       // strip everything (re-encode)
await privakuTools.run("remove-photo-location", file, { remove: ["gps", "c2pa"] })   // lossless, only those groups (JPEG/PNG)
r.blob, r.name, r.before.gps, r.after.gps
await privakuTools.run("screenshot-to-text", { files: [img], lang: "eng+jpn" })  // [{ text, confidence }]
await privakuTools.run("reduce-photo-size", { files: [img], maxEdge: 1280, quality: 0.7, maxKB: 200 })
await privakuTools.run("image-to-pdf", { files: [a, b, c], name: "receipts" })    // [{ blob, pages }]
await privakuTools.run("photo-privacy-check", { files: [img], blurFaces: true, shareCard: true })
// [{ score, level, leaks, location, device, date, faces, aiSignals, blob (clean copy), card }]

URL parameters

/tools/reduce-photo-size/?maxEdge=1280&quality=0.7&maxKB=200
/tools/screenshot-to-text/?lang=jpn
/tools/remove-photo-location/?remove=gps,xmp
/tools/photo-privacy-check/?blurFaces=true

Parameters preset the options; the file is still chosen in the page or passed to run().

WebMCP

In browsers that support WebMCP (an early proposal, in a Chrome origin trial), each tool page registers its tool with document.modelContext.registerTool. Other browsers ignore it.

npm package and MCP server (coming soon)

@privaku/tools (Node) and privaku-tools-mcp (MCP over stdio) run the same tools on the agent's own machine with no network access: inspect_metadata, remove_metadata, photo_privacy_check, screenshot_to_text, reduce_photo_size, image_to_pdf. Files are read from and written to local paths. Face detection is web-only for now. They are not published yet, so there is no install command today.

Fair use

There is no server, API key or limit. Please don't wrap these tools in a service that uploads other people's files.

All tools · Privaku apps · Home