Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    your token

    1

    Track Claude token usage, forecast API costs, and automate project handoffs before hitting usage limits.

    Free

    1 useSecurity scanned
    your token

    your token

    Example session with this skill installed

    I'm about to start a massive refactor of this 20-file repo. Check my current token status and let me know if I have enough room to finish or if we should hand off now.

    • Read your context and instructions
    • Compiled the your token

    I have checked the current ledger. You are currently at 42% of your 5-hour window usage and 15% of the conversation context limit. A 20-file refactor involves high input costs due to repeated context re-reads. ## Token Usage Report | Metric | Current Value | Limit/Budget | Status | | :--- | :--- | :--- | :--- | | **Window Usage** | 12,600 units | 30,000 units | OK | | **Context Fill** | 31,450 tokens | 200,000 tokens | OK | | **Forecasted Cost** | ~18,500 units | Next 10 turns | WARN | | **Model** | Sonnet 3.5 | Plan: Pro | - | Based on the forecast, a deep refactor of 20 files will likely trigger a `HANDOFF` event within 8-10 turns because each tool call re-reads the context. **Next steps** 1. Run `python scripts/ledger.py predict --effort 10 --tool-calls 20` to get a precise trajectory. 2. If you want to proceed, I will log every turn and alert you at the 70% warning threshold. 3. Consider setting `--read-mode skim` for the initial file scan to preserve your budget.

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

    About this skill

    Avoid hitting the dreaded "Usage limit reached" message mid-task. Large-scale coding tasks and long research sessions often exhaust Claude's 5-hour window or API budget without warning. This skill provides a local ledger to track token consumption, forecast costs, and trigger clean handoffs before you lose access. ## What it does - **Usage tracking** monitors input, output, and hidden thinking tokens for both claude.ai and API modes. - **Budget forecasting** predicts when a task will hit limits based on reasoning effort and tool call frequency. - **Automated handoffs** generates a `HANDOFF.md` state file and saves all in-progress work when limits reach 85%. - **Cost estimation** calculates marginal costs for the next message and total USD spend for API users. - **Plan calibration** adjusts for Free, Pro, and Team plan multipliers or custom API rate limits. ## How it works 1. **Initialization** starts the ledger using `init` for API modes or `plan --set` for claude.ai subscriptions. 2. **Logging** records message turns, file reads, and tool results using the `turn` and `add` commands. 3. **Monitoring** checks the status at 70% (WARN) and 85% (HANDOFF) thresholds to prevent data loss. 4. **Resuming** restores the ledger state in a new chat using the `restore` command from a handoff file. ## Frameworks & tools Built with Python 3 using only the standard library. Operates locally via a JSON ledger at `~/.your-token/ledger.json` and requires a code execution environment for precision tracking. ## Why this beats prompting it yourself Manual estimates rarely account for the recursive cost of context re-reads or the specific pricing ratios of different Claude models. This skill uses precise multipliers for thinking tokens and cache discounts that are impossible to calculate mentally. ## Use cases - Managing long coding refactors that require dozens of tool calls and deep file reads. - Controlling API spend by setting hard USD caps and monitoring real-time usage objects. - Transitioning large projects across conversation context limits without losing progress. - Optimizing reasoning effort settings to stretch a limited 5-hour window. ## Known limitations Cannot see real-time server-side limits from Anthropic. Estimates for claude.ai are based on verified plan multipliers but require calibration if the baseline free-tier changes.

    Cannot access real-time server-side usage data directly. claude.ai limits are based on community-derived baselines and require manual calibration for absolute accuracy.

    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 any payment.

    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

    1 use

    Times used to date

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 15 days ago

    • Free with an Agensi account

    Requires

    Python

    Permissions

    Terminal / Shell
    Read Files
    Write Files
    Network Access
    Environment Variables
    Listed15 days ago

    What's inside

    Frequently Asked Questions

    This skill prevents data loss and interrupted workflows by tracking your token consumption locally and automatically generating a project state handoff file before you hit Anthropic's usage limits.

    The skill is built in Python and runs in any environment with code execution capabilities; it is compatible with Claude 3.5 Sonnet, Opus, and Haiku via both the web interface and API.

    Setup is instantaneous—simply run the 'init' command to start a ledger and 'plan --set' to calibrate the tracker to your specific Free, Pro, or Team subscription tier.

    No, the ledger is stored locally at '~/.your-token/ledger.json' and does not require an external database or internet connection, ensuring your usage data and project summaries remain private.

    You receive the full skill package, including the Python-based tracking logic, a documentation guide for command usage, and life-time access to any version updates as Anthropic adjusts their token pricing or plan limits.

    While the skill cannot see internal Anthropic servers, it uses high-precision multipliers for reasoning tokens and cache hits to provide an estimate that is significantly more accurate than manual guessing or basic word counts.

    New in Productivity