Loading ToolBench
Official MCP server, REST API & CLI for Social Neuron — create, schedule, and optimize social content with 20+ AI models across YouTube, TikTok, and more.
The Social Neuron MCP server has 25 tools with significant structural issues. Most tools have descriptions, but they are often marketing-focused rather than operationally clear. Parameter schemas are present but incomplete, many lack proper type constraints and validation guidance. A critical pattern violation: tool names like 'plan_content_week' and 'plan_content_calendar' are nearly identical, creating disambiguation risk for LLMs. Schema documentation is minimal; output structures are not explicitly documented. Error handling is absent from visible code. The server exposes high-risk destructive operations (delete_content_plan, schedule_post, create_autopilot_config) without confirmation or dry-run patterns. Naming conventions are inconsistent (noun_adjective vs verb_noun). Parameter descriptions are sparse, most lack type constraints, ranges, or format guidance. No evidence of idempotency markers or pagination helpers for list operations.
Check the status of an async job (video generation, post scheduling, etc.) and retrieve results.
Create an autopilot automation rule to automatically generate, optimize, and schedule content based on triggers and templates.
Delete a content plan and all associated posts.
Detect unusual performance patterns and anomalies in post analytics.
Execute a predefined workflow recipe (multi-step automation) like 'weekly calendar + schedule' or 'trend detection + ideate + create'.
Extract brand identity, voice, audience, and messaging from a URL or company website.
Generate a performance summary and insights for recent posts across platforms.
CRITICAL: 'plan_content_week' and 'plan_content_calendar' are ambiguous sibling tools that differ only in granularity. LLMs cannot distinguish when to call which. Both names lack a clear action verb and differ from standard verb_noun conventions.
CRITICAL: Destructive operations (schedule_post, delete_content_plan, create_autopilot_config, respond_plan_approval, execute_recipe) lack confirmation patterns, dry-run options, or reversibility. Agents can irreversibly schedule posts or delete plans without safeguards.
HIGH: Parameter descriptions are marketing-oriented, not operationally precise. Example: quality_check's platform description is 'Target platform (instagram, youtube, tiktok, linkedin)', no format constraint, no enum declaration visible in schema, no guidance on what happens if platform is invalid.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 48 | 2025-06-18+ | v2 |
| 2026-04-15 | F | 27 | - | v1 |
Generate a short-form video using AI with a storyboard, music, voiceover, and brand customization.
Retrieve platform analytics (views, engagement, reach, impressions) for a project or post.
Retrieve a content plan with AI-generated posts, platform-specific optimization, and scheduling recommendations.
Get current credit balance and usage statistics.
Generate AI content ideas for a given topic with platform-specific hooks, angles, and engagement strategies.
Retrieve AI-optimized ideation context for a project - top hooks, winning patterns, recommended models, and ideal posting times based on historical performance.
List all active autopilot automations for a project.
List comments on a post with sentiment analysis, platform, and user details.
List pending approvals for content plans with draft/approved status.
Create a multi-week content calendar with AI-generated posts, scheduling recommendations, and engagement strategies.
Generate the plan
Check out our new AI tool! #ai #tech
Approve or reject a content plan, optionally with edits.
Generate and post an AI response to a comment, with optional brand voice customization.
Save or update a brand profile with custom voice, audience, colors, and messaging guidelines.
Schedule a post to be published at a specific time on one or more platforms.
Search and discover MCP tools by name, module, or keyword. Returns descriptions and usage guidance for agent decision-making.
Take a screenshot of a URL or web content for visual reference, preview, or quality checking.
HIGH: Input schemas lack comprehensive type information. Most parameters are typed as 'string' or 'array' without minLength, maxLength, pattern, or enum constraints. An LLM calling 'schedule_post' with an invalid ISO 8601 timestamp will fail without guidance on the correct format.
HIGH: Output schemas are not documented in visible code. Tools like 'generate_performance_digest' and 'detect_anomalies' have READ_ONLY risk but no documented return structure. LLMs cannot plan downstream tool calls without knowing what fields to expect.
HIGH: List operations (list_comments, list_autopilot_configs, list_plan_approvals) have no pagination parameters (page, limit, offset, cursor). Unbounded lists risk context window exhaustion and token waste.
MEDIUM: Parameter naming inconsistencies. Some tools accept 'project_id', others accept 'plan_id' or 'approval_id'. Chaining tools requires the LLM to map outputs to inputs, if a tool returns 'plan_id' but the next expects 'plan_id' vs 'id', errors cascade.
MEDIUM: No error recovery guidance. Visible tool descriptions do not include recovery hints (e.g., 'If the post fails to schedule, call check_status(job_id) to diagnose, or contact support'). Agents lack direction on what to do when a call fails.
MEDIUM: Tool 'respond_to_comment' and similar tools accept tone parameters (friendly, professional, humorous) as free-form strings without enum constraints. LLMs may invent invalid tones like 'sarcastic' or 'corporate'; no validation visible.
MEDIUM: Inconsistent naming patterns. Some tools use underscores for clarity (quality_check, search_tools), others combine concepts (ideation_context, autopilot_config). No consistent verb_noun or noun_adjective convention makes it harder for LLMs to parse intent from names.