Loading ToolBench
MCP server for Grok Imagine image generation and editing via xAI API
Server has 6 tools with complete JSON schemas and descriptions. Naming follows verb_noun convention (configure_, generate_, edit_, continue_, get_, list_). Descriptions are present and adequate (50-150 chars). However, parameter descriptions lack depth, many omit usage context, constraints, and error recovery guidance. Output schemas are not documented. Error handling is minimal; no recovery guidance or actionable error messages. Security concern: configure_xai_token accepts API keys as parameters (should use env var injection). No tool annotations (readOnlyHint, destructiveHint). Composition is reasonable but continue_editing relies on session state (lastImagePath), which is fragile.
Configure your xAI API token for Grok Imagine image generation
Continue editing the LAST image that was generated or edited in this session without specifying a file path.
Edit a SPECIFIC existing image file, optionally using up to 2 additional reference images. Use this when you have the exact file path of an image to modify.
Generate a NEW image from a text prompt using Grok Imagine models.
Check the current authentication status and configuration of the server.
List all previously generated and edited images stored locally.
API credentials exposed as tool parameter. configure_xai_token accepts apiKey as input parameter, which will be logged in traces and prompt history. Should use server-side secret injection via environment variables.
No output schemas documented. Tools return results but LLMs cannot predict response structure. generate_image, edit_image, and continue_editing should document returned fields (e.g., image_url, image_path, generation_id).
Stateful session handling (lastImagePath). continue_editing relies on global lastImagePath variable, making it fragile and non-idempotent across sessions. Should accept explicit image_path parameter or return session context.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 64 | <=2025-11-25 | v2 |
Minimal error handling. No recovery guidance in error responses. When API calls fail, errors are not categorized as retryable vs. fatal. LLM receives no actionable next steps.
No tool annotations. Tools lack readOnlyHint (get_configuration_status, list_generated_images) and destructiveHint (generate_image, edit_image consume API credits). Annotations help LLMs reason about side effects.