Loading ToolBench
A Next.js-based AI video generation and editing studio that integrates with Anthropic's Claude, MuAPI (Seedance), and Atlas Cloud for video generation, editing, and composition. Provides an agentic interface for filmmakers to generate, edit, and assemble video content.
This server has 11 tools with significant definition gaps. Most tools lack proper input schemas in the source code, only render_edit and generate_video show complete JSON Schema definitions. Descriptions are present but generic (50-150 chars), lacking LLM-optimized guidance on WHEN to use each tool or what happens on failure. Parameter descriptions are sparse or missing entirely. Tools like edit_video, extend_video, and atlas_generate_video are inferred from ToolCard.tsx rather than explicitly registered, capping their scores at 50. No output schemas are documented. Error handling is absent, no recovery guidance, no categorization of retryable vs fatal errors. The server conflates video generation, editing, and assembly into separate tools without clear composition guidance.
Check Atlas Cloud account balance and available credits.
Check the status and results of an Atlas Cloud video generation prediction.
Generate video using Atlas Cloud models with support for various model architectures and parameters.
Apply edit passes to an existing video take, including color grading, effects, and style adjustments.
Extend an existing video take by generating additional frames to lengthen its duration.
Generate a continuation of an existing video, extending the narrative or action.
Generate a video from text prompt using various routes (text-to-video, image-to-video, first-last-frame, omni-reference). Supports aspect ratio, duration, resolution, seed, and variant selection.
9 of 11 tools lack visible input schemas in source code. Only render_edit and generate_video show complete JSON Schema definitions. Tools like edit_video, extend_video, atlas_generate_video are inferred from UI components, not explicitly registered with schemas.
Descriptions are generic and lack LLM-optimized guidance. Most descriptions (30-50 chars) do not explain WHEN to use the tool, what happens on failure, or how it differs from similar tools. E.g., edit_video says 'Apply edit passes' without explaining what 'edit passes' means or when to call it vs generate_video.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 36 | 2026-07-28+ | v2 |
Generate voiceover audio from text using text-to-speech synthesis.
Perform QA checks on a generated video take, including technical validation and quality assessment.
Assemble completed canvas takes into ONE finished edit (trim, transitions, captions, voiceover) and save it as a downloadable MP4. Use whenever the producer asks for a cut, a version, a social edit, or 'put it together'. clips reference canvas takes by videoJobId in order; start/end are optional trim seconds inside each take (omit for full length — you know each take's durationSec). transition: 'cut' or an xfade name (fade, wipeleft, wiperight, slideup, slidedown, circleopen, dissolve, fadeblack, zoomin, smoothleft). texts are burned captions with timeline seconds. audioAssetId picks a workspace voiceover/music asset. Returns a downloadUrl the producer can click.
Generate a style-matched reshoot of a video take with consistent visual characteristics.
No output schemas documented. Agents cannot plan downstream calls or extract required fields. render_edit mentions 'downloadUrl' but no structured response schema is defined. Other tools have no documented return types.
No error handling or recovery guidance. Tools lack descriptions of failure modes, retryable vs fatal errors, or what the LLM should do next. E.g., atlas_check_prediction has no guidance on what to do if a prediction fails or times out.
Tool composition is unclear. Multiple tools generate or edit video (generate_video, edit_video, extend_video, atlas_generate_video, reshoot_matching, generate_continuation) without clear guidance on which to use when. No tool chains documented.