A comprehensive MCP server for publishing content to social media platforms (Twitter, YouTube, Instagram) with analytics, scheduling, content optimization, and account insights capabilities
Server has 10 tools with complete JSON schemas and descriptions. Naming follows verb_noun convention (publish_post, get_analytics, etc.). However, descriptions are verbose (200-400+ chars, exceeding the 10-200 char sweet spot for LLM parsing), parameter descriptions lack actionable constraints (e.g., no min/max for limits, no format specs for timestamps), and output schemas are entirely undocumented. Error handling and recovery guidance are absent. No tool annotations (readOnlyHint, destructiveHint, idempotentHint) despite clear risk levels. Idempotency keys are present but not enforced in schema validation.
Cancel a scheduled post before its publication time
Generate an OAuth 2.0 PKCE authorization URL to connect a user's social media account
Retrieve account-level analytics, follower growth, total reach/impressions, profile activity, recent posts performance comparison, and algorithmic health diagnosis to identify why reach or engagement is low
Retrieve engagement metrics (impressions, likes, retweets, replies) for a published post
List all scheduled posts with their metadata, status, and scheduled publication times
Generate high-converting algorithmic SEO optimization (viral hooks, high-search keyword tags, structured descriptions, and niche hashtag clusters) and optionally apply updates to live posts with user permission
Output schemas completely undocumented. No tool returns a documented response structure. LLMs cannot plan downstream calls or extract required fields (e.g., post_id from publish_post, scheduled_post_id from schedule_post). This violates pattern:tool and forces agents to guess response shapes.
Descriptions are verbose (200-400+ chars) and lack actionable constraints. E.g., 'Optional prompt to auto-generate a stunning AI image visual on-the-fly' is marketing copy, not a constraint. Missing: format specs for timestamps (ISO 8601 not stated), min/max for limit parameter (default 20 but no bounds), enum values for time_period in get_account_insights (day|week|days_28|lifetime not in description).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | <=2025-11-25 | v2 |
Healthcheck and MCP protocol connection handshake verification
Publish a text, image, or video post (via public URL or Base64 media_data) to a connected social platform with idempotency protection
Schedule an autonomous social media post, Instagram Reel, YouTube Video/Short, or Tweet for future publication with AI copy generation support
Upload and stage AI-generated images or videos into a public HTTPS URL accessible by Instagram/YouTube/Twitter crawlers
No tool annotations despite clear risk levels. publish_post and schedule_post are WRITE operations; optimize_content_seo with apply_update=true is destructive; cancel_scheduled_post is REVERSIBLE. Missing readOnlyHint, destructiveHint, idempotentHint annotations prevent agents from reasoning about side effects and retry safety.
Error handling and recovery guidance absent. No tool documents what errors can occur, whether they are retryable, or what the LLM should do next. E.g., if publish_post fails due to invalid platform, should the agent retry, ask the user, or call connect_platform first? No guidance.
Parameter descriptions lack format and constraint details. E.g., scheduled_time is 'ISO 8601 timestamp' in description but no pattern or example. media_data is 'Base64-encoded binary string' but no size limits. time_period enum values (day|week|days_28|lifetime) are in schema but not described in parameter text, forcing LLMs to read JSON Schema.