Loading ToolBench
MCP Server for Speakly speech recognition providing tools for audio transcription and provider management
The Speakly MCP server defines 2 tools with explicit schemas and basic descriptions. However, the definitions have significant gaps in parameter documentation, output schema clarity, and error handling guidance. Tool naming is adequate (verb_noun pattern), but descriptions lack context on when to use each tool and what to expect in responses. Parameter descriptions are minimal, 'Provider to use (optional, uses default if not specified)' tells the LLM nothing about which providers are available or what format is expected. The output schema is not documented anywhere in the visible source, leaving the LLM to guess what fields transcribe_audio_file returns. No error handling guidance exists, if transcription fails, the agent has no recovery path. The tools are READ_ONLY, which is appropriate, but no such hint is embedded in the tool definition (toolAnnotations absent). Overall, these tools are functional but underspecified for LLM use.
List available speech recognition providers
Transcribe an audio file to text
Output schemas not documented. LLMs cannot anticipate the structure or fields returned by transcribe_audio_file or list_providers, forcing them to guess or make follow-up calls.
Parameter descriptions are too brief and vague. 'Provider to use (optional, uses default if not specified)' tells the LLM nothing about which providers exist, their names, or format. 'Language code (e.g., 'zh', 'en', 'auto')' includes example values (anti-pattern); should use enum or format constraint instead.
No error handling guidance. If transcribe_audio_file fails (file not found, unsupported format, provider unavailable), the LLM receives only an error code or message with no hint on recovery actions (retry, fallback provider, adjust language, validate file path).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 34 | 2024-11-05+ | v1 |
Missing tool annotations. The tools are READ_ONLY (safe, idempotent), but no readOnlyHint or idempotentHint is declared in the schema. Agents cannot determine tool safety or retry behavior without these hints.
No dependency documentation. list_providers should hint that agents often call it first to discover options before invoking transcribe_audio_file. transcribe_audio_file should document that it requires a provider name from list_providers (or default behavior).