io.github.dontpayfull/ainotate

What it mainly works on, as our AI model read its tool descriptions: files and code on the machine it runs on; a web browser or desktop applications.

7 of 16 tools are described as changing something (creating, editing, deleting, sending or running). For 2 the reading was uncertain, so no mark is shown.

• annotateon 7 other serverscreates or changes datadescribed as changing data

  “Draw marks on an existing image: numbered steps, boxes, arrows, labels, highlights, spotlight, text and solid redaction.”

  Publisher declares: read-only no · destructive no · idempotent no · open-world no

  3 inputs, 1 required

  • specobject or stringrequired“Render spec as a JSON object (or a JSON string).”
  • draftboolean“true (default): write to a temp folder so you can check the preview first.”
  • lookstring or null“Mark style: "default" (bold, built in) or the name of an installed look package (e.g.”
• previewon 9 other serversreads or searches

  “Debug render of an annotate spec: draws the marks plus cyan target rects and magenta label boxes so you can see why a label moved or a mark misses its element.”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  2 inputs, 1 required

  • specobject or stringrequired“Render spec as a JSON object (or a JSON string).”
  • lookstring or null“Mark style: "default" (bold, built in) or the name of an installed look package (e.g.”
• capture_webreads or searches

  “Screenshot a web page in a headless browser (Playwright) and measure named elements, so you can annotate with exact rects.”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  11 inputs

  • urlstring or null“Page to load (http/https).”
  • cdp_urlstring or null“DevTools address of a running browser to attach to, e.g.”
  • page_url_containsstring or null“With cdp_url: substring of the URL of the tab to capture.”
  • presetstring“Viewport: "laptop" 1280x800@2, "wide" 1600x1000@2 or "phone" 390x844@3.”
  • targetsobject or null“Named elements to measure: {"name": {"text": "Checkout"} | {"selector": "css"} | {"role": "button", "name": "Save"}}, each with optional "within", "nth" (0-bas…”
  • actionsarray or null“Steps run in the page before the capture, in order: {"click": T}, {"fill": T, "value": "..."}, {"wait": T}, {"scroll_to": T}, {"wait_ms": 500}; T is a target a…”
  • hide_overlaysboolean“Hide cookie banners and similar overlays before the capture and after each click.”
  • full_pageboolean“Capture the whole scrollable page instead of the viewport.”
  • timeout_msnumber“Time limit in ms for loading and for each action or target.”
  • browserstring“Engine to launch: "chromium" (default), "firefox" or "webkit".”
  • user_data_dirstring or null“Browser profile folder to reuse (cookies, logins) when launching.”
• shootcreates or changes datadescribed as changing data

  “One call from URL to annotated image: capture_web + annotate, with marks that point at elements by "target" instead of "rect".”

  Publisher declares: read-only no · destructive no · idempotent no · open-world yes

  3 inputs, 1 required

  • specobject or stringrequired“Render spec as a JSON object (or a JSON string).”
  • draftboolean“true (default): write to a temp folder so you can check the preview first.”
  • lookstring or null“Mark style: "default" (bold, built in) or the name of an installed look package (e.g.”
• locateon 8 other serversreads or searches

  “Find visible text on an image by OCR and return its rect, for images where you have no DOM or accessibility data (desktop captures, the user's screenshots).”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  8 inputs, 2 required

  • image_pathstringrequired“Path of a PNG, JPEG or WebP image on this machine (~ allowed).”
  • textstringrequired“Visible text to find: a word or short phrase as it appears on the image.”
  • allboolean“Also list weaker matches, not only the best ones.”
  • nthinteger or null“Pick one of several equal matches, 0-based in reading order.”
  • withinarray or null“Search only inside [x1, y1, x2, y2], image px.”
  • exactboolean“Require the exact text (case and spacing), not a close match.”
  • langstring or null“OCR language code, e.g.”
  • backendstring or null“OCR engine: "vision" (macOS), "winrt" (Windows), "tesseract", or omit for auto.”
• capture_screenreads or searches

  “Screenshot the whole screen or one region, as the user sees it right now.”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  1 input

  • regionarray or string or null“Part of the screen: [x, y, w, h] in screen points, or the same four numbers as a string "x,y,w,h".”
• list_windowson 8 other serversreads or searches

  “List the visible windows: [{id, app, title, bounds}], to get the id for capture_window.”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  1 input

  • filterstring or null“Case-insensitive substring of the app name or window title.”
• capture_windowreads or searches

  “Screenshot one window by id (from list_windows); never moves, resizes or focuses it.”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  2 inputs, 1 required

  • idinteger or stringrequired“Window id from list_windows.”
  • targetsobject or null“macOS only.”
• clipboard_image

  “Save the image that is on the clipboard (e.g.”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  Takes no inputs.

• gridon 1 other servercreates or changes datadescribed as changing data

  “Overlay a labelled coordinate grid every `step` px on an image, to find roughly where an element is when locate cannot read it (icons, images, no text).”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  2 inputs, 1 required

  • image_pathstringrequired“Path of a PNG, JPEG or WebP image on this machine (~ allowed).”
  • stepinteger“Grid spacing in image px (> 0).”
• zoom

  “Enlarged crop of [x1,y1,x2,y2] with a fine grid labelled in SOURCE image px, so you can read an exact rect straight off the preview (spec scale 1).”

  Publisher declares: read-only yes · destructive no · idempotent yes · open-world no

  6 inputs, 5 required

  • image_pathstringrequired“Path of a PNG, JPEG or WebP image on this machine (~ allowed).”
  • x1numberrequired“Left edge of the region, source image px.”
  • y1numberrequired“Top edge of the region, source image px.”
  • x2numberrequired“Right edge of the region, source image px (greater than x1).”
  • y2numberrequired“Bottom edge of the region, source image px (greater than y1).”
  • stepinteger“Fine grid spacing in source px (> 0).”
• make_guidedescribed as changing data

  “Build a step-by-step guide from annotated images: Markdown, HTML and/or PDF, one image and text per step, every image at its own size.”

  Publisher declares: read-only no · destructive no · idempotent no · open-world no

  5 inputs, 1 required

  • stepsarrayrequired“Guide steps in order: [{"image": path, "text": "...", "title"?: "...", "alt"?: "..."}].”
  • titlestring“Guide title.”
  • formatsarray or null“Any of "md", "html", "pdf".”
  • namestring“Base file name in the output folder, no extension.”
  • draftboolean“true (default): write to a temp folder so you can check the preview first.”

Each quoted sentence is the first sentence of the publisher’s own description. The kind of tool and the “changes data” mark are our AI model’s reading of that description, not something we tested: two differently worded runs had to agree, otherwise no label is shown, and nobody has checked them by hand. A tool without the “changes data” mark is described as only reading, searching or calculating, or the two runs did not agree. No tool was called. Read the publisher’s full description before relying on it. “Publisher declares” lists the hints the server itself sends with the tool (MCP annotations: read-only, destructive, idempotent, open-world); they are the publisher's claims, not checked by us. Inputs are the parameters in the tool's own input schema: name, declared type, whether it is required, and the first sentence of its description. Read 2026-10-23.

The registry verifies a package once, when it is published, and does not re-check it. A difference below is about the listing, not a judgement of the publisher.