Agent Optimization — AI skill for Claude Code

Agent Optimization

[](https://codeguilds.dev/packages/agent-optimization)

**Live token & cost monitor for Claude Code and Codex sessions.** Tracks every API call you make, finds what's wasting money, and lets you fix it — automatically.

---

What is this?

If you use Claude Code or Codex heavily, your token spend compounds fast — long sessions, large context files, wrong models for the job, keepalive probes eating cache writes. This dashboard reads your local session files in real time, quantifies the waste, and gives you actionable fixes.

Built for people who ship with AI agents every day and want to stay in control of the cost.

---

Features

📊 Overview — Live token dashboard

Real-time breakdown of every token you've consumed across Claude Code and Codex sessions.

• 5 KPI cards — Total tokens · Input · Output · Cache Read · Cache Create, each with USD cost
• Agent split — Claude Code vs Codex side-by-side bar charts with per-category cost
• Token composition — Stacked bar showing exactly where your budget goes
• Session table — Filterable by source, sorted by recency, with per-session cost breakdown

🗂 Projects — Project-level drill-down

Group sessions by project. Click into any project to see every session that ran inside it.

• Cost and token totals per project, sorted by spend
• Cache hit ratio, output share, avg cost/session
• Per-session table with session name (first user message), model, all token buckets + cost

🔬 Analysis — LLM-powered waste detection

Every 2 hours, Claude Haiku analyses your projects using your existing OAuth session — **no API key needed**.

**Detected patterns:**

• 🔴 Context bloat — loading huge context for minimal output
• 🔴 Cache inefficiency — writing cache blocks that never get reused
• 🟡 Overpowered model — using Opus on tasks Sonnet would handle equally well
• 🟡 Output-heavy sessions — verb