Loading ToolBench
Professional MCP server for annotating screenshots with markers, arrows, callouts, and more. Works seamlessly with Playwright MCP for documentation workflows.
Server has 4 tools with complete input schemas and descriptions. annotate_screenshot has excellent, detailed description (500+ chars) with premium tips and color/theme enums. However, output schemas are entirely undocumented, LLMs cannot plan downstream calls or extract results. get_image_dimensions lacks parameter descriptions. create_step_guide and highlight_area have minimal descriptions (under 100 chars). No error handling guidance. No tool annotations (readOnlyHint/destructiveHint). Naming is clear (verb_noun pattern). Parameters have types but some lack descriptions (e.g., highlight_area's 'shape' enum is documented but 'x', 'y', 'width', 'height' lack context on coordinate system or units).
Add professional annotations to a screenshot image. Annotation types available: • marker - Numbered circles (1, 2, 3...) with gradient and shadow • arrow - Straight arrows with customizable heads • curved-arrow - Smooth curved arrows • callout - Text boxes with pointers (speech bubbles) • rect - Rectangle highlights • circle - Circle highlights • label - Text labels with optional backgrounds • highlight - Semi-transparent overlays • blur - Blur sensitive content • connector - Dashed lines between elements • icon - Icon badges (check, x, warning, info, question) • spotlight - Dim the whole screenshot and softly highlight one area (place FIRST) Themes: documentation, tutorial, bugReport, highlight Colors: red, orange, yellow, green, blue, purple, pink, cyan, teal, white, black, gray, lightGray, darkGray, success, warning, error, info, primary, secondary, accent PREMIUM TIPS for clean docs: • Markers auto-size to the screenshot resolution (retina-aware). Pass "scale" to override. • To avoid covering the UI, give a marker a "target": [x,y] of the element and let it sit in a clear area — it draws a leader line to the target automatically. • Set "margin" (e.g. 60) to pad the canvas so edge callouts/labels are never clipped. • Add a "spotlight" annotation FIRST to dim everything but the focus area. • Set "frame": { "browserBar": true } for a rounded browser chrome + drop shadow. • "format": "webp" + "downscale": 0.6 give crisp, light images for docs sites. • Text now uses a clean professional font by default (no handwriting).
Create a numbered step-by-step guide on a screenshot. Automatically places numbered markers with labels and connecting arrows. Perfect for tutorials and documentation.
Get width, height, and format of an image. Essential for calculating annotation coordinates.
Output schemas completely undocumented. LLMs cannot infer what annotate_screenshot, create_step_guide, or highlight_area return. No field names, types, or structure documented.
get_image_dimensions parameter 'image_path' lacks description. LLMs cannot infer whether it expects absolute or relative paths, or what happens if the file doesn't exist.
create_step_guide and highlight_area have minimal descriptions (45 - 55 chars). Too short to guide LLM selection. Missing WHEN to use, WHAT it returns, and prerequisites.
No error handling guidance. Tools that write files (annotate_screenshot, create_step_guide, highlight_area) offer no recovery hints if output_path is invalid, disk is full, or input image is corrupted.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 60 | 2026-07-28+ | v2 |
Quickly highlight a specific area with a shape and optional label.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot distinguish read-only get_image_dimensions from write-heavy annotate_screenshot without explicit hints.