MCP server for Scenario.com API - provides tools for image generation, asset management, and creative AI workflows
This server has 3 tools with mixed quality. Tool naming follows verb_noun convention (set-, get-), but descriptions vary significantly in clarity and actionability. The set-credentials tool has clear, well-scoped input with appropriate descriptions. The get-credentials-status tool is minimal but functional. The get-assets tool has extensive parameters but with critical gaps: many parameters lack proper descriptions (e.g., 'sortBy', 'pageSize', 'originalAssets' are minimally documented), and the tool lacks any documented output schema. Parameters like 'pageSize' have vague descriptions ('The number of items to return') without stating the default or practical range beyond the enum definition. Error handling is not visible in the source code provided. Tool descriptions are adequate but not LLM-optimized, they lack context on when to use each tool or what the return structure looks like. The server appears to auto-generate tools from Swagger/OpenAPI (via scripts/generate_tools.js), which explains the inconsistent documentation quality; generated descriptions are often copied verbatim from API docs without LLM adaptation.
List assets of a project team, or list all public assets
Check if credentials are configured for this session (does not expose actual values).
Configure your Scenario.com API credentials for this session. You MUST call this tool first before using any other Scenario tools. Get your credentials from https://app.scenario.com/ (Settings → API Keys).
get-assets has no documented output schema. LLMs cannot plan downstream calls or extract required fields (e.g., asset IDs for chaining) without knowing the response structure.
get-assets parameter descriptions are incomplete or vague. 'pageSize' says 'The number of items to return' but omits the default (50) and practical guidance. 'sortBy' lists values but not when each is useful. 'originalAssets' has no description at all.
No error handling guidance visible. If API calls fail, the handler returns raw axios errors without categorizing them as retryable, user-fixable, or fatal. LLMs receive stack traces instead of actionable recovery steps.
get-credentials-status accepts no parameters but has minimal description ('Check if credentials are configured'). The description should explain what this returns (boolean? status object?) and when to call it.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | C | 64 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
Credentials are stored in session variables (sessionCredentials global), but no tool explicitly documents the session lifetime, what happens on timeout, or whether credentials persist across MCP client reconnections. This is a hidden dependency.