Loading ToolBench
Headless AI promo video capture, composition, rendering, editing, and review toolkit
This server has 40 tools with severe definition gaps. Tool #6 (list_music) has description 'E', a single character. Many tools lack visible input schemas in the provided source excerpt. Descriptions vary wildly: some are detailed (e.g., help, inspect_saas), others are single words or fragments. Parameter documentation is sparse, most tools show only parameter names without type info or constraints visible in the code sample. No evidence of enum constraints, format declarations, or validation guidance. Error handling descriptions are absent. The server exposes file paths and spec files as parameters without clear validation or sanitization guidance. Output schemas are not documented in the visible code.
Add selected motion component source groups to the composition.
Analyze audio file for tempo, key, energy, and instrumentation to match motion pacing.
Run an authenticated Playwright asset plan and save clean product screenshots and/or application recordings plus a manifest.
Prepare final delivery output by preserving license metadata, removing review artifacts, and validating completeness.
Create a comprehensive visual review package with project, motion plan, render variables, and frame sheets.
Download free stock image or graphic with license metadata and attribution.
Tool #6 (list_music) has description 'E', a single character. This violates the 10 - 1024 character requirement and provides zero guidance to LLMs on when or why to call this tool.
Input schemas are not visible in the provided source code for most tools. The code excerpt shows tool names and descriptions but does not display the actual schema registration (e.g., z.object() definitions or parameter type declarations). This makes it impossible to verify parameter types, constraints, or descriptions.
File path parameters (specPath, screenshotPath) are exposed without documented validation or sanitization. No evidence of path traversal guards, file type validation, or size limits. Tools like capture_saas and record_saas_flows accept arbitrary file paths.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | <=2025-11-25 | v2 |
Download free stock video with license metadata and attribution.
Download licensed music from Openverse with metadata and attribution.
Edit image with crop, scale, rotate, or color adjustments.
Edit video range, trim, or composite multiple clips with transitions.
Extract key frames from video for visual review and quality assessment.
Generate or update the bundled audio library for offline use.
Progressively load detailed documentation for one exact MCP tool, motion component, transition, or production topic. Call without arguments for a compact index, with query for short candidates, then with target kind:id (or kind plus id) for complete parameters, values, constraints, examples, pitfalls, and validation. For tolerance, query values already shaped as kind:id are promoted automatically to exact target lookups.
Inspect a SaaS page for title, messaging, headings, dominant colors, and an optional screenshot.
List all scaffolded Revideo projects in the workspace.
List all available motion component source groups that can be added.
List all available motion components organized by category.
E
List all supported video format profiles (landscape, portrait, square) and platform targets.
Load the complete create-ai-promo-video Skill workflow before planning a production. Use this first when the host client exposes MCP tools but does not discover filesystem Skills, prompts, or resources, such as Claude Desktop Home.
Mix multiple audio tracks with crossfades, volume envelopes, and timing.
Describe the motion engine capabilities and constraints.
Apply targeted edits to TypeScript composition or motion-plan.json without full rewrite.
Prepare caption timing from speech audio with word-level precision or cue-level interpolation.
Probe video file for duration, dimensions, codec, audio presence, and silence detection.
Read the complete TypeScript composition and motion-plan.json from a scaffolded project.
Record authenticated application interactions as clean, high-quality MP4 clips controlled by the AI. Supports click, fill, press, hover, select, check, scroll, mouse movement, navigation, and timing.
Render a scaffolded Revideo composition to MP4 with optional worker parallelization.
Render a video from a JSON specification file using the legacy JSON renderer.
Replace a time range in a video with another clip, preserving audio or video as specified.
Resolve a video format ID and platform target to dimensions and safe area insets.
Review settled caption frames and word timing before final render.
Save authored TypeScript composition and motion-plan.json back to disk.
Create a minimal Revideo composition with motion-plan.json and TypeScript scaffolding.
Search for free stock images and graphics from Pexels, Pixabay, Openverse, and Wikimedia.
Search for free stock video footage from Pexels, Pixabay, Openverse, and Wikimedia.
Search the motion component library by category, name, or keyword.
Search for music by mood, genre, tempo, or keyword across Freesound, Openverse, and other licensed sources.
Validate motion-plan.json for completeness, boundary declarations, and focal velocity preservation.
Validate a capture or render specification file for required fields and format compliance.
No error handling guidance visible. Tools that perform destructive operations (render, edit, delete) do not document recovery steps, retryability, or confirmation patterns. Agents cannot determine whether to retry or escalate.
Output schemas are not documented. Tools return complex objects (video metadata, motion plans, frame sheets) but the response structure is not visible in the provided code. LLMs cannot plan downstream tool calls without knowing what fields are returned.