Loading ToolBench
An MCP server that plans work against your Claude session quota. Reads quota snapshots and spend data to derive burn rate and pack tasks into remaining capacity.
quota-planner demonstrates solid tool design with clear naming, comprehensive descriptions, and well-structured schemas. All 7 tools follow verb_noun conventions (plan_task, record_task_cost, status, reconcile_spend, set_budget, task_history, reset_costs). Descriptions are detailed (100-250 chars) and explain WHAT, WHEN, and consequences. Input schemas are complete with types and descriptions for all parameters. Tool annotations (readOnlyHint, destructiveHint) are correctly applied. However, output schemas are not documented in the source, LLMs cannot see what fields to expect from responses. Error handling is minimal; no recovery guidance or actionable error messages visible. Parameter constraints (enums for model/category/mode) are well-defined but lack format/range documentation in descriptions.
Plan a task against remaining quota or budget. Returns whether the task fits, how much quota/spend it will consume, and how much remains after.
Fetch billed spend from Anthropic Admin API and reconcile against local estimates. Requires ANTHROPIC_ADMIN_KEY or ANTHROPIC_ADMIN_OAUTH_TOKEN.
Record the actual cost of a completed task to refine future estimates. Updates the seed estimates with measured data.
Clear all recorded task costs and revert to seed estimates.
Set daily and/or monthly USD budget for spend-mode planning. Persists to budget.json.
Get current quota/spend status: remaining capacity, burn rate, time until reset, and mode (quota vs spend).
Retrieve recorded task costs from history. Returns up to N most recent entries with model, category, cost, and timestamp.
Output schemas not documented. LLMs cannot see what fields plan_task, status, task_history, or reconcile_spend return. This forces agents to guess structure and risks parsing failures.
No error handling guidance. Tools like reconcile_spend (requires Admin API credential) and reset_costs (destructive) lack recovery hints. If reconcile_spend fails due to missing credential, the LLM has no guidance on what to do next.
reset_costs is destructive but lacks confirmation/dry-run pattern. Agents can irreversibly clear all task cost history without a safety gate.
Parameter descriptions lack format/range details. 'days' in reconcile_spend has no min/max; 'limit' in task_history has no bounds. LLMs may pass absurd values (days=999999, limit=1000000).
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 62 | 2026-07-28+ | v2 |
reconcile_spend requires Admin API credential but does not document which environment variables are checked or what happens if none are set. Error message is in code but not visible to LLM.