Permissionless communication supercharger MCP server — 40+ Lightning-paid tools: AI phone calls in any language, voice in 602 languages, translation across 119 plus 452 rare languages frontier models don't serve, fax, SMS, transcription, audiobooks, and more. No signup, no API keys, no KYC.
Critical failures across the board. The source code provided contains NO explicit tool schemas, parameter descriptions, or input/output documentation. Tools are inferred from package.json keywords and index.js mentions only, but actual schema definitions are not visible in the provided code sample. The descriptions provided in the evaluation input (e.g., 'Generate images using AI models') are generic, under 50 characters, and lack the specificity required by the rubric. No parameter documentation is evident. No error handling guidance is visible. This server appears to be a thin wrapper around external APIs with minimal intrinsic tool documentation. The check-server-json.js script validates registry metadata drift but does nothing to validate tool schemas or descriptions.
Generate 3D models using AI
Make AI-powered phone calls in any language
Make phone calls
Check the status of a job
Check the status of a payment
Colorize black and white images
Convert HTML to PDF
NO INPUT SCHEMAS VISIBLE: Not a single tool in the provided source code shows explicit parameter schemas, types, or constraints. Tools cannot be used reliably without agents knowing what parameters they accept and what types are expected.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 36 | 2026-07-28+ | v2 |
Create a payment for services
Deblur images
Detect NSFW content in images
Detect objects in images
Send emails
Convert EPUB files to audiobooks
Extract data from receipts
Convert files between formats
Get the result of a completed job
Get pricing information for models
Generate images using AI models
Edit images using AI
List available AI models
List planned services
Generate music using AI models
Extract text from documents using OCR
Merge PDF files
Remove background from images
Remove objects from images
Render boarding pass or wallet cards
Request a refund for a service
Restore faces in images
Send SMS messages
Generate text using AI models
Transcribe audio to text
Translate text across 119 plus 452 rare languages
Text-to-speech conversion supporting 602 languages
Upscale images to higher resolution
Generate videos using AI models
Generate videos from images using AI models
Analyze images using vision AI
Clone voices for speech synthesis
Vote on service features or improvements
GENERIC, SHORT DESCRIPTIONS: Tool descriptions like 'Generate images using AI models', 'Analyze images using vision AI', and 'Extract text from documents using OCR' are under 50 characters, lack specifics about parameters, prerequisites, return values, or when to use each tool instead of similar ones. LLMs cannot make informed tool selection decisions on these.
NO PARAMETER DESCRIPTIONS: The code sample shows no visible parameter documentation anywhere. No hints on what fields agents should pass, what types are expected, what constraints apply, or what happens if required params are missing.
NO DOCUMENTED OUTPUT SCHEMAS: It is unclear what structure agents should expect from any tool's response. No indication of field names, types, whether results are paginated, what IDs are returned for chaining, or what metadata is included.
NO ERROR HANDLING DOCUMENTATION: No visible guidance on what errors tools can return, whether they are retryable, whether they require user intervention, or what the LLM should do next on failure. This violates the recovery-guide and error-classification patterns.
INFERRED TOOL DEFINITIONS: Tool existence is inferred from package.json keywords and index.js mentions, but no explicit tool registration code is visible in the provided source. However, since NO schemas are visible at all, they all score 0 for schema and under 30 overall.
DESTRUCTIVE OPERATIONS WITHOUT CONFIRMATION: Many tools (create_payment, call, send email, sms) are marked WRITE but no source shows confirmation or dry-run patterns. Irreversible operations should support a confirmation step to prevent accidental charges or messages.
UNBOUNDED LIST RESULTS: Tools like list_models and list_planned_services have no visible pagination, limit, or offset parameters. Without these, large result sets will bloat context windows and degrade LLM reasoning.
LACK OF TOOL ANNOTATIONS: The MCP spec supports toolAnnotations (readOnlyHint, destructiveHint, idempotentHint) to guide agent planning. None are visible in the source, forcing LLMs to infer safety properties from descriptions alone (which are too generic to infer from).