A quantified self tracking system with MCP integration for logging and querying food, expenses, time, and body measurements
Life Tracker MCP has solid naming conventions (verb_noun pattern across all 13 tools) and comprehensive parameter descriptions with format constraints (YYYY-MM-DD regex, ISO 8601 for time). However, output schemas are entirely undocumented, tools return JSON.stringify() with no declared structure, forcing LLMs to parse unstructured text. Error handling is minimal (only log_body_measurement shows error classification). Tool descriptions are good (avg ~150 chars) but lack dependency hints and recovery guidance. No tool annotations (readOnlyHint/destructiveHint) despite clear read/write distinction. Batch operations (log_expense, log_meal, log_time) are well-designed but lack pagination/limit documentation for retrieval tools.
Get daily-averaged body measurements for a metric over a date range.
Get expense summary for a date range, broken down by day and by category.
Get food and nutrition summary for a date range. Returns daily totals and per-meal breakdowns with macro nutrients.
Get time tracking summary for a date range, broken down by bucket.
Get today's date. Always call this first to avoid date errors.
List all body measurement metrics the user has defined (e.g. Weight, Bicep). Always call this before logging a measurement to get valid metric IDs.
Output schemas completely undocumented. All retrieval tools (list_body_metrics, get_body_measurements, get_expense_summary, get_food_summary, get_time_summary, list_money_categories, list_wallets, list_buckets) return JSON.stringify() with no declared field structure. LLMs cannot plan downstream calls or extract required IDs without parsing unstructured text.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Tools are marked READ_ONLY or WRITE in metadata but not exposed via MCP tool annotations. Agents cannot infer safety/retry semantics from the tool definition alone.
Inferred effective spec: 2026-07-28+.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 67 | 2026-07-28+ | v2 |
List all time buckets for the current user (id, name, color, is_distraction).
List all money categories for the user. Returns id, name, display_label, color, and type ('expense' or 'income'). Call this before log_expense or log_income to resolve the correct category_id.
List all active wallets with their IDs, names, types, and current balances. Call this before log_expense to resolve the correct wallet_id.
Log a body measurement for a specific metric. Call list_body_metrics first to get valid metric IDs. The user cannot log to a metric that doesn't exist.
Log one or more expenses in a single call. IMPORTANT: Always call list_money_categories first to get valid category IDs for type 'expense'. Always call list_wallets first to get valid wallet IDs. Never omit wallet_id — always debit the correct wallet.
Log one or more food items for a meal. Each item in a meal is logged separately but grouped by meal_type.
Log one or more time blocks in a single call. Use full ISO datetimes so cross-midnight activities (like sleep) are handled correctly. Use list_buckets to get valid bucket IDs.
Retrieval tools lack pagination/limit documentation. get_body_measurements, get_expense_summary, get_food_summary, get_time_summary return unbounded results with no mention of limits or pagination. Large result sets will exhaust context windows.
Error handling minimal. Only log_body_measurement shows error classification (isError flag). Other write tools (log_expense, log_meal, log_time) lack recovery guidance. Agents cannot distinguish retryable from fatal errors.
Dependency hints missing. Tool descriptions mention 'Call list_body_metrics first' and 'Call list_wallets first' but lack explicit guidance on when/why. Descriptions should state: 'Always call X before Y to resolve IDs' to prevent wasted calls.