Niyantra provides 13 read-only tools with clear, domain-specific descriptions (avg ~150 chars). All tools have descriptions and most parameters are typed. However, critical gaps undermine production readiness: (1) No input schemas visible for 6 tools (quota_status, usage_intelligence, budget_forecast, analyze_spending, switch_recommendation, codex_status, quota_forecast, copilot_status, plugin_status), these accept empty objects or no parameters, but schema definitions are not shown in source. (2) Parameter descriptions are present but minimal (10-30 chars), lacking format constraints, ranges, or validation guidance. (3) No output schemas documented, LLMs cannot predict response structure for chaining. (4) No error handling guidance, tools return data or fail silently. (5) Tool names are descriptive but lack verb-noun consistency (e.g., 'quota_status' vs 'get_quota_status'). All tools are READ_ONLY, reducing risk, but lack of structured output documentation and error recovery patterns limits agent reliability.

• Document output schemas for all 13 tools. Specify field names, types, and structure. Example: quota_status should return {accounts: [{name: string, quota_remaining: number, reset_time: string}], ...}. This enables LLM chaining and reduces hallucination.
• Add explicit input schemas to tools with optional/empty parameters. Use JSON Schema with type, description, enum, minimum, maximum, and pattern constraints. Example: token_usage_stats should have {days: {type: integer, minimum: 1, maximum: 365, default: 30}, provider: {type: string, enum: [all, claude, antigravity, codex, cursor, gemini]}}.
• Expand parameter descriptions to 50-100 chars with format/constraint details. Current: 'number of days to analyze (default 30, max 365)'. Better: 'Number of days of historical data to analyze. Range: 1 - 365 days. Default: 30. Larger windows may be slower.'
• Implement error handling with recovery guidance. When a tool fails (e.g., insufficient data for rate calculation in usage_intelligence), return {error: string, suggestion: string}. Example: {error: 'Insufficient tracking data', suggestion: 'Wait 30+ minutes for rate calculations to stabilize, then retry.'}
• Rename tools to verb_noun pattern for LLM clarity. quota_status → get_quota_status, model_availability → check_model_availability, usage_intelligence → get_usage_intelligence. This aligns with 90% of production tools and improves LLM intent inference.

Inferred effective spec: <=2025-11-25.

• Relies on Stateful initialize / Mcp-Session-Id (removed; protocol is stateless) - make each request self-contained