Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Codex & Claude Code Quota Capacity Estimator

    1

    Estimate full 5-hour and weekly Codex or Claude Code quota capacity in tokens and API-equivalent USD from supplied aggregates, with clear coverage limits and observed usage shown separately.

    Free

    Security scanned
    Codex & Claude Code Quota Capacity Estimator

    Codex & Claude Code Quota Capacity Estimator

    Example session with this skill installed

    Use quota-capacity-estimator to calculate a 5-hour full-capacity reference from these synthetic teaching aggregates. Run the supplied calculator, keep observed usage separate from the extrapolation, and return its full JSON with an explanation. These example values and prices are not measurements of a real subscription.

    {
      "samples": [
        {
          "provider": "codex",
          "account": "local-codex-a",
          "window": "5h",
          "start": "2026-11-12T00:00:00Z",
          "resets_at": "2026-11-12T05:00:00Z",
          "as_of": "2026-11-12T02:00:00Z",
          "used_percent": 25,
          "usage": {
            "account": "local-codex-a",
            "start": "2026-11-12T01:00:00Z",
            "as_of": "2026-11-12T02:00:00Z",
            "coverage": "partial",
            "subscription_only": true,
            "total_tokens": 10000000,
            "api_equivalent_usd": 20,
            "price_basis": "Synthetic worked-example cost only; not a real provider rate."
          },
          "calibration": {
            "baseline": {
              "provider": "codex",
              "account": "local-codex-a",
              "window": "5h",
              "start": "2026-11-12T00:00:00Z",
              "resets_at": "2026-11-12T05:00:00Z",
              "as_of": "2026-11-12T01:00:00Z",
              "used_percent": 10
            },
            "identity_stable": true,
            "limits_clear": true
          }
        }
      ]
    }
    
    • Read your context and instructions
    • Compiled the codex & claude

    Result: observed-scope reference

    This is a synthetic worked example executed with the v1.1.0 calculator, not a live account reading. The quota rises from 10% to 25%, a 15-percentage-point increase. The supplied interval contains 10,000,000 tokens and $20 of API-equivalent usage.

    MeasureObserved intervalFull-window reference
    Tokens10,000,00066,666,667
    API-equivalent USD$20.00$133.33

    Calculation: 10,000,000 / 0.15 = 66,666,667 tokens after rounding; $20 / 0.15 = $133.333333. Because coverage is partial, this remains an observed-scope reference, not a verified account-wide capacity, remaining balance, or guaranteed bound.

    Complete calculator output

    {
      "estimates": [
        {
          "provider": "codex",
          "account": "local-codex-a",
          "window": "5h",
          "as_of": "2026-11-12T02:00:00Z",
          "status": "reference",
          "full_tokens": 66666667,
          "full_api_equivalent_usd": 133.333333,
          "method": "delta",
          "observed": {
            "start": "2026-11-12T01:00:00Z",
            "as_of": "2026-11-12T02:00:00Z",
            "total_tokens": 10000000,
            "api_equivalent_usd": 20,
            "coverage": "partial",
            "subscription_only": true,
            "price_basis": "Synthetic worked-example cost only; not a real provider rate."
          },
          "reasons": [],
          "warnings": [
            "Partial/unknown coverage: observed-scope reference only; missing usage can bias this value. Not account-wide capacity or a guaranteed bound.",
            "Linear scenario at the supplied as_of, not an official capacity or remaining balance.",
            "USD is API-price equivalence, not subscription charges, credit or cash value."
          ]
        }
      ]
    }
    

    A weekly estimate needs its own matched window and usage sample. Missing price evidence would leave USD unavailable; missing identity or payment evidence would block extrapolation. To evaluate predictive accuracy, obtain a separate valid later interval and compare its observed quota change with a prediction from the earlier interval. This example establishes arithmetic only.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Estimate full 5-hour and weekly quota capacity for Codex and Claude Code from account-matched usage aggregates. This focused, read-only calculator keeps observed usage separate from conditional capacity estimates.

    What it does

    • Calculates whole-window scenarios or estimates from the usage and percentage increase between two observations in the same window.
    • Reports tokens and API-equivalent USD independently for each account and each window.
    • Labels results as estimated (complete supplied coverage), reference (partial or unknown coverage), or unavailable (insufficient or conflicting evidence).
    • Checks supplied account aliases, timestamps, coverage and payment flags. It does not authenticate accounts or independently verify those flags.

    How to use it

    1. Provide sanitized JSON aggregates following the included input contract: opaque account alias, quota window and used percentage, matched token totals, and API-equivalent cost with its price basis.
    2. Ask your agent to run the standalone Python calculator.
    3. Review observed usage, full-capacity scenarios, and the reasons or warnings attached to each result.

    For delta calibration, supply direct interval usage and two quota observations from the same account and reset window. Account stability, exclusion of extra paid usage, and clear quota boundaries must be established by your data source.

    Included

    A self-contained Python 3.9+ calculator, input and method documentation, worked examples for Codex, Claude Code, Remote magpie exports and delta calibration, plus an MIT license. No third-party Python dependencies or built-in price feed.

    Scope and limitations

    This skill calculates from supplied aggregates. It does not collect logs, connect to Magpie or provider accounts, monitor quotas, or provide a dashboard. Remote magpie is an accepted source of correctly prepared aggregates, not an automatic integration.

    Estimates assume the observed usage mix continues. Missing logs, caching, model changes, percentage rounding and provider reporting delays can bias results. Partial-coverage references are not account-wide capacities or guaranteed bounds. USD is API-price equivalence, not actual subscription charges, included cash credit or refundable balance. Remaining tokens and exhaustion-time forecasts are not produced.

    Arithmetic and execution have been tested; real-world capacity prediction accuracy has not been established. This independent tool is not an official OpenAI, Anthropic or Magpie quota specification. Version 1.1.0. MIT licensed.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for LLM Ops.

    User-supplied aggregates only; no automatic account authentication, log collection, Magpie connection, monitoring or dashboard. Only fixed 5-hour and 168-hour windows are supported. Input account, coverage and paid-usage flags must be substantiated externally. Partial coverage produces a reference, not account-wide capacity. Unknown prices suppress USD; uncertain paid usage, account/time mismatches, exhaustion and insufficient percentage changes block extrapolation. Exact reset timestamps must match for delta calibration; timestamp jitter may reject a segment. No remaining-token or exhaustion-time prediction. Model mix, caching, rounded percentages and delayed reporting can bias estimates. Real-world prediction accuracy is not yet established. USD means API-price equivalence, not actual spending or cash credit.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before buying a skill.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free to use

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 5 days ago

    • Free with an Agensi account

    Software
    Python 3.9+

    Supply sanitized quota and usage aggregates in the documented JSON format. Python runs locally. No provider login, API key or Magpie connection is required. Account attribution, coverage and applicable prices must be established before calculation.

    Permissions

    Terminal / Shell
    Read Files
    Write Files

    Read access is limited to the delivered skill files and the aggregate input explicitly supplied for the task. When delivered as text via MCP, the agent writes the unchanged calculator to a task temporary file, executes it, then removes it. The calculator itself only reads stdin and writes stdout/stderr. No credential stores, native account logs, routing/login settings, billing APIs or network services are accessed or modified. Do not provide emails, keys, raw account IDs or conversation contents.

    Listed5 days ago

    What's inside

    Frequently Asked Questions

    No. It extrapolates from supplied usage and quota percentages. Complete supplied coverage yields an estimated scenario; partial or unknown coverage yields an observed-scope reference. Neither is an official capacity or a guaranteed lower bound. It does not forecast remaining tokens or time to exhaustion.

    Use an agent that can run Python 3.9+ locally. The calculator has been executed with Codex CLI and Claude Code. It accepts Codex or Claude Code aggregates, including manually prepared Remote magpie exports, only when they match the included JSON contract. It does not accept arbitrary raw logs or automatically connect to those services.

    One SKILL.md, one standalone Python calculator, two method/example references, four synthetic JSON examples and an MIT license: nine files in version 1.1.0. There are no third-party Python dependencies, built-in API keys or live price feed.

    The calculator reads sanitized aggregate JSON from stdin and writes results to stdout. It does not access credentials, native logs, billing APIs or the network. Your agent may materialize the script in a temporary file and read the input you provide. Use opaque account aliases and exclude emails, keys and conversation contents.

    Supply the API-equivalent cost of the same requests, calculated externally using applicable model, input/output, cache and tier rates, with a price source and date. The skill extrapolates that aggregate; it does not fetch prices or price raw tokens itself. Missing cost or price evidence leaves USD unavailable. API-equivalent USD is not the subscription bill or cash credit.

    The v1.1.0 calculator passed 17 behavior tests and both Codex CLI and Claude Code executed it on sanitized real-account aggregates with matching outputs. This verifies arithmetic and execution, not real-world prediction accuracy. Account/time mismatches, uncertain extra paid usage, exhausted quotas, zero observed tokens, or less than 5 percentage points of signal block extrapolation. Delta calibration also requires stable identity and clear quota boundaries.

    New in LLM Ops

    Free