Read Claude Code sessions

    1

    See which model actually answered, where the tokens went, and what compaction silently threw away.

    Free

    Security scanned

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Read Claude Code sessions

    Example session with this skill installed

    Why did this session cost so much, and did it switch models on me?

    • Read your context and instructions
    • Compiled the read claude code

    Summary

    Requested claude-sonnet, answered by claude-opus (substitution recorded at turn 3).
    Tokens: 412k input, 38k output, 351k cache reads.
    Compactions: 2 - discarded 118k tokens at turn 41.
    Upstream errors: 1 x HTTP 529, retried once, recovered.

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

    About this skill

    Claude Code writes a complete record of every session to disk and then shows almost none of it. This skill reads it.

    It reports: the model the client asked for next to the model that answered (a substitution is a fact in the log, not a guess); the full token and cache accounting; every context compaction with the tokens it discarded; every upstream error with its retry counter.

    Three things it refuses to do, because each is a way to look right while being wrong:

    • It never reports a guess as a measurement. If the log shape is unfamiliar, it says unmeasured and stops.
    • It never calls the cost figure a bill. The client's own pricing table marks unknown models, and that is surfaced.
    • It never says "it forgot" without the compaction record as evidence.

    Read-only. Node 20+. No dependencies, no network, no telemetry.

    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.

    Requires Node.js >= 20. Only reads local Claude Code logs. Cannot modify transcripts or fix upstream billing issues.

    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 2 days ago

    • Free with an Agensi account

    Software
    Node.js

    Permissions

    Terminal / Shell
    Read Files
    Listed2 days ago

    What's inside

    Frequently Asked Questions

    This skill addresses 'hidden' issues like silent model substitutions during peak traffic, context window compaction that causes memory loss, and invisible API retries that lead to unexpected latency.

    The skill is a Node.js utility that requires local read access to your `~/.claude/projects/` directory; it works with any version of Claude Code that stores session data in this standard path.

    No, this is a read-only auditing tool that parses existing JSONL log files on your machine to provide a forensic report; it does not modify your Claude Code configuration or project files.

    You will receive the Node.js source files, a CLI entry point for running audits, and documentation on interpreting the 'Verdict' reports generated by the tool.

    The skill requires Node.js 20 or higher and needs permission to read files within your home directory's hidden Claude folder to access the raw telemetry data.

    Yes, you can use the `--since` flag to filter by time or the `--session` flag with an ID prefix to audit a specific interaction, allowing you to focus on one problematic bug or billing spike.

    New in LLM Ops

    Free