Loading ToolBench
This server demonstrates good definition quality with well-structured tool schemas, clear descriptions, and proper naming conventions. All 8 tools follow verb_noun patterns (sora_list_*, sora_get_*, sora_generate_*) and include substantive descriptions (150-400 chars each). Input schemas are complete with proper types and enums for constrained parameters. However, there are gaps in output documentation, limited error handling guidance, and missing tool annotations. The tools follow composition patterns well, with clear separation of concerns between generation, status checking, and information retrieval. Descriptions helpfully indicate when to use each variant (text-to-video vs image-to-video vs character-based). Parameters use enums effectively for model, size, duration, and orientation. The main weakness is lack of documented output schemas, the code shows these tools return Task objects with specific fields, but the tool definitions don't document what fields agents should expect, forcing LLMs to infer structure.
Generate an AI video from a text prompt using Sora. This is the primary way to create videos - describe what you want and Sora will generate a video matching your description. Use this when: - You want to generate a video from a text description - You don't have reference images - You want creative AI-generated video content For image-to-video generation, use sora_generate_video_from_image instead. For character-based video generation, use sora_generate_video_with_character. Returns: Task ID and generated video information including URLs and state.
Generate an AI video asynchronously with callback notification. This is useful for long-running video generation tasks. Instead of waiting for the video to complete, you'll receive a callback at your specified URL when the generation is finished. Use this when: - You don't want to wait for the generation to complete - You have a webhook endpoint to receive results - You're integrating with an async workflow The callback will receive a POST request with the same response format as the synchronous generation tools. Returns: Task ID that you can use to correlate with the callback.
Generate an AI video from reference images using Sora (Image-to-Video). This allows you to animate or create videos based on provided images. The AI will use the images as visual references for the generated video. Use this when: - You have reference images you want to animate - You want the video to match a specific visual style - You want to bring static images to life Returns: Task ID and generated video information including URLs and state.
Output schemas not documented. Tools return Task objects with status, video URLs, and metadata, but tool definitions don't specify what fields are in the response. LLMs must infer structure from context or fail to chain tool calls effectively.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Tools like sora_get_task and sora_get_tasks_batch should be marked readOnly=true. Async generation tools should be marked as having side effects. Missing annotations force LLMs to reason about safety without protocol guidance.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | B | 78 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 17 | - | v1 |
Generate an AI video featuring a character from a reference video. This allows you to create new videos featuring a specific character extracted from another video. The character will be placed in the new scene described by the prompt. IMPORTANT: The reference video must NOT contain real people. Only animated or digital characters are supported. Use this when: - You want to reuse a character in different scenes - You're creating a series with the same character - You want consistent character appearance across videos Returns: Task ID and generated video information including URLs and state.
Query the status and result of a video generation task. Use this to check if a generation is complete and retrieve the resulting video URLs and metadata. Use this when: - You want to check if a generation has completed - You need to retrieve video URLs from a previous generation - You want to get the full details of a generated video Task states: - 'pending': Generation is still in progress - 'succeeded': Generation finished successfully - 'failed': Generation failed (check error message) Returns: Task status and generated video information including URLs and state.
Query multiple video generation tasks at once. Efficiently check the status of multiple tasks in a single request. More efficient than calling sora_get_task multiple times. Use this when: - You have multiple pending generations to check - You want to get status of several videos at once - You're tracking a batch of generations Returns: Status and video information for all queried tasks.
List all available Sora API actions and corresponding tools. Reference guide for what each action does and which tool to use. Helpful for understanding the full capabilities of the Sora MCP. Returns: Categorized list of all actions and their corresponding tools.
List all available Sora models and their capabilities. Shows all available model versions with their limits, features, and recommended use cases. Use this to understand which model to choose for your video generation. Returns: Table of all models with their version, limits, and features.
Error handling lacks recovery guidance. No documented error cases, categories (retryable vs user-fixable vs fatal), or suggested recovery actions. Example: if API rate limit is hit, should the agent retry? How long to wait? What's the user-visible error?
Pagination and result limits not addressed. sora_get_tasks_batch accepts a list of task IDs but mentions 'maximum recommended batch size is 50' only in description. No pagination guidance for hypothetical future bulk listing tools. No explicit cap on returned results documented.
Parameter constraints could be more explicit. 'character_start' and 'character_end' accept 0-1 range but description doesn't specify minimum, maximum, or whether inclusive/exclusive. 'task_id' has no format hints (UUID? alphanumeric?). Reduces clarity for LLM input validation.