More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    jev pilot

    1

    Optimize Claude Code costs by automatically selecting reasoning effort and subagent models per prompt.

    Free

    1 useSecurity scanned
    jev pilot

    jev pilot

    Example session with this skill installed

    How do I install jev-pilot to save on my Claude Code token usage?

    • Read your context and instructions
    • Compiled the jev pilot

    I can help you set up jev-pilot to optimize your Claude Code usage limits and costs. It uses a fast decision model to pick the right reasoning effort and subagent for each task.

    Installation Instructions

    To get started, you will need an OpenRouter API key. Once you have it, run these commands in your terminal:

    
    # 1. Add and install the plugin
    claude plugin marketplace add Akramovic1/jev-pilot
    claude plugin install jev-pilot@jev-pilot --config openrouterApiKey=<your_key_here> --config timeoutMs=1500
    
    # 2. Enable function hooks in your environment
    export CLAUDE_CODE_ENABLE_FUNCTION_HOOKS=1
    
    # 3. Launch Claude Code
    claude
    

    Alternatively, you can make the function hooks permanent by adding them to your ~/.claude/settings.json:

    {
      "env": {
        "CLAUDE_CODE_ENABLE_FUNCTION_HOOKS": "1"
      }
    }
    

    Once running, you will see a status indicator above your prompt. You can use /jev to see active switches or /jev-pilot:report to view recent optimization data.

    Next steps

    • Generate an API key at openrouter.ai if you haven't already.
    • Run the installation commands provided above.
    • Verify the ready · openrouter status appears in your next Claude Code session.

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

    About this skill

    Running Claude Code on xhigh reasoning for every task quickly exhausts usage limits and inflates costs. This skill integrates jev-pilot to automate model selection and reasoning effort, ensuring you only pay for Opus when the task actually requires it.

    What it does

    • Automates reasoning levels by setting effort from low to xhigh based on task complexity.
    • Optimizes subagent selection by routing simple lookups to Haiku and well-defined coding tasks to Sonnet.
    • Handles failed tool calls by automatically raising reasoning effort when a subagent struggles.
    • Smart skill attachment by identifying and attaching only the specific SKILL.md required for the current prompt.
    • Usage reporting via /jev-pilot:report to visualize decision records and tuning suggestions.

    How it works

    1. Installation of the jev-pilot plugin via the Claude Code marketplace using an OpenRouter API key.
    2. Environment configuration to enable function hooks in Claude Code settings or via the claude-jev launcher.
    3. Automatic classification where a fast decision model analyzes each prompt to set parameters before execution.
    4. Command-line management using /jev to toggle features or check status mid-session.

    Frameworks & tools

    This skill is purpose-built for Claude Code (CLI) version 2.1.278 or newer. It requires an OpenRouter API key to access the Jev decision model and utilizes Claude Code function hooks for real-time intervention.

    Why this beats prompting it yourself

    Manually switching models or effort levels for every command is friction-heavy and prone to human error. This skill automates the decision logic at the hook level, reducing token waste without requiring the user to think about the underlying infrastructure.

    Use cases

    • Cost reduction when running high-volume terminal tasks that don't all need Opus.
    • Rate limit management by offloading read-only operations to faster, lighter models.
    • Workflow optimization by automatically attaching relevant skills to the context.

    Known limitations

    Requires Claude Code 2.1.278+. Needs an OpenRouter API key. Only prompt text and metadata are sent to the classifier, not full code files. Function hooks must be enabled.

    Requires Claude Code 2.1.278+. Needs an OpenRouter API key. Classifier only sees prompt and metadata, not full file contents. Function hooks must be manually enabled.

    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 1 month ago

    • Free with an Agensi account

    Permissions

    Terminal / Shell
    Write Files
    Listed1 month ago

    What's inside

    Frequently Asked Questions

    jev pilot acts as an automated routing layer for Claude Code, intelligently adjusting reasoning levels (low to xhigh) and selecting the most cost-effective subagent model for each specific prompt. This prevents unnecessary Opus usage and token waste on simple tasks like file lookups or basic refactors.

    This skill is purpose-built for Claude Code (CLI) version 2.1.278 or newer. It requires an OpenRouter API key to access the decision model and must be run in an environment where Claude Code function hooks are enabled.

    Setup involves installing the plugin via the marketplace and configuring your OpenRouter API key. Once installed, you can launch using the `claude-jev` command or enable function hooks in your existing Claude Code settings for automatic intervention.

    No, privacy is a priority. The classifier only analyzes the prompt text and metadata to determine the required reasoning effort; it does not send your full codebase or sensitive files to the decision model.

    The skill includes the jev-pilot plugin, the `claude-jev` launcher, and access to the `/jev-pilot:report` command for usage visualization. Buyers also receive documentation on configuring smart skill attachments for automated context management.

    New in LLM Ops