Claude Trip Computer

Version 0.13.7 | Changelog

Real-time session analytics and optimization insights for Claude Code. TypeScript-powered with multi-line status display, git integration, tool activity tracking, and efficiency metrics.

Quick Setup

Prerequisites

Node.js 18+ - Check: node --version (Download)
Claude Code - Latest version recommended

Automated Installation (Recommended)

Windows:

# Double-click or run:
install.bat

Linux/macOS:

./install.sh

The installer will:

✓ Check Node.js version (18+ required)
✓ Detect and remove old bash-based installations automatically
✓ Prompt for billing mode (API or Subscription)
✓ Create configuration files
✓ Update Claude Code settings
✓ Create /trip command
✓ Test the installation

Time: ~2 minutes | Then restart Claude Code

Upgrading from bash version? The installer automatically detects and removes old bash scripts, hooks, and cache files. No manual cleanup needed!

Manual Installation

Click to expand manual setup instructions

Clone repository:

cd ~/Code  # or your preferred location
git clone https://github.com/luxsolari/claude-trip-computer

Configure billing mode:

Create ~/.claude/hooks/.stats-config:

# Claude Code Session Stats Configuration
BILLING_MODE="API"  # or "Sub" for subscription
BILLING_ICON="💳"   # or "📅" for subscription
SAFETY_MARGIN="1.00"  # 1.10 for subscription (10% buffer)

Configure Claude Code status line:

Edit ~/.claude/settings.json:

{
  "statusLine": {
    "type": "command",
    "command": "npx -y tsx /full/path/to/claude-trip-computer/src/index.ts"
  }
}

Replace /full/path/to/ with your actual path.

Restart Claude Code

Time: ~5 minutes

Need Help?

Troubleshooting → TROUBLESHOOTING.md
Technical Docs → CLAUDE.md
Want to contribute? → CONTRIBUTING.md

What You Get

Status Line

Real-time session efficiency metrics in your Claude Code status bar:

Multi-Line Status (v0.13.2):

💬 5 msgs (Opus 4.5) | 🔧 12 tools (2.4/msg) | 🎯 15.3K tok | 🌿 main* | ████░░░░░░ 35% | ⚡ 78% cached | 📝 3.1K/msg | ⏱️ 31m | 📈 /trip
✓ Edit ×55 | ✓ Bash ×41 | ✓ Read ×23 | ✓ Write ×8
✓ Explore: Explore codebase structure (52s)
▸ Fix authentication bug (2/5)

Line 1 - Session Metrics:

💬 5 msgs (Opus 4.5) - Message count with current model
🔧 12 tools (2.4/msg) - Tool usage with intensity ratio
🎯 15.3K tok - Total tokens (deduplicated)
🌿 main* - Git branch with dirty indicator (NEW)
████░░░░░░ 35% - Context window usage bar
⚡ 78% cached - Cache efficiency
📝 3.1K/msg - Response verbosity
⏱️ 31m - Session duration (NEW)
📈 /trip - Trip computer link
📅 ~$X.XX value (Sub users only) - API-equivalent value

Line 2 - Tool Activity (NEW):

Top 5 tools by frequency with accurate session-wide counts
Running tools: ◐ Bash: command...

Lines 3+ - Agent Status (NEW, when agents exist):

Shows running/completed agents with duration

Final Line - Todo Progress (NEW, when todos exist):

In progress: ▸ Task name (2/5)
Complete: ✓ All todos complete (5/5)

Trip Computer

Ask Claude to show trip computer stats, or run directly:

npx tsx /path/to/claude-trip-computer/src/index.ts --trip-computer

Session optimization dashboard with complete billing-mode differentiation:

For API Users - Optimization-First Experience:

═══════════════════════════════════════════════════════════════
  📊 TRIP COMPUTER - Session Analytics Dashboard
═══════════════════════════════════════════════════════════════

📊 QUICK SUMMARY
───────────────────────────────────────────────────────────────
  Health: ⭐⭐⭐⭐⭐ Excellent (85/100)
  Messages: 3 | Tools: 46 | Tokens: 3.8M

📈 SESSION HEALTH (0-100)
───────────────────────────────────────────────────────────────
  Overall: ⭐⭐⭐⭐⭐ 85/100

  ⚡ Cache Efficiency: ✅ 40/40 points
     90% cache hit rate

  ⚙️  Context Management: ➡️ 15/30 points
     Context tracking unavailable

  🎯 Efficiency: ✅ 30/30 points
     15.3 tools/msg, 4.4K tok/msg

🤖 MODEL MIX
───────────────────────────────────────────────────────────────
  Sonnet 4.5
  █████████░ 92.7% of cost
  Tokens: 3.5M

  Haiku 4.5
  ░░░░░░░░░░ 7.3% of cost
  Tokens: 336.7K

📊 TOKEN DISTRIBUTION
───────────────────────────────────────────────────────────────
  Input: 0.1% ░░░░░░░░░░
  Output: 0.3% ░░░░░░░░░░
  Cache writes: 9.6% ░░░░░░░░░░
  Cache reads: 90.0% █████████░

⚡ EFFICIENCY METRICS
───────────────────────────────────────────────────────────────
  Tool Intensity: Minimal - early session or simple tasks
    46 tools (15.3 tools/msg) across 3 msgs

  Response Verbosity: Concise - brief responses
    4.4K tokens/msg average

  Output/Input Ratio: 3.13x

  Cache Hit Rate: 90.4%
    Excellent → stay in session

📊 SESSION METRICS
───────────────────────────────────────────────────────────────
  Messages: 3 | Tools: 46
  Cache Efficiency: 90.4%
  Total Tokens: 3,815,792

  ℹ️  Use /cost for official billing amounts

🎯 TOP OPTIMIZATION ACTIONS
───────────────────────────────────────────────────────────────
  ✅ Session looks well-optimized! Keep up the good work.

📊 SESSION INSIGHTS
───────────────────────────────────────────────────────────────
  Context Growth: Slow growth → healthy pace
    4.4K tokens/msg average

  Tool Pattern: Minimal - early session or simple tasks
    46 tools (15.3 tools/msg)

  Cache Performance: Excellent → stay in session
    90.4% hit rate

For Subscription Users - Value + Optimization:

Same dashboard with 💵 Cost Drivers (dollar amounts)
📈 Trajectory section with projected costs
Dollar savings in optimization recommendations
API-equivalent estimates with 10% safety margin

How It Works

Technical Implementation

Architecture: TypeScript executed via npx tsx (no compilation needed)

Claude Code Features Used:

Status Line Command (~/.claude/settings.json)
- Not a hook - direct command execution by Claude Code
- Invoked automatically on every interaction
- Receives JSON stdin: {session_id, model, context_window, ...}
- Returns formatted string to status bar
Custom Slash Command (~/.claude/commands/trip.md)
- User-invocable skill via /trip
- Assistant executes command with --trip-computer flag
- Assistant displays output in code block (instructed by skill docs)
- No hooks involved - pure command execution
No Hooks Used (v0.11.0+)
- Old bash version (<0.11.0) used SessionEnd hooks for automatic session-end stats
- v0.11.0 removed all hooks - uses only status line command + slash command
- Installer automatically removes old SessionEnd hooks from settings.json

Status Line Flow:

User interaction → Claude Code invokes status line command
→ Script reads stdin (session_id, model, context)
→ Parses transcript: ~/.claude/projects/<project>/<session>.jsonl
→ Deduplicates tokens by requestId + model
→ Aggregates per-model, calculates costs
→ Returns: "💬 X msgs (Model) | 🔧 X tools | ..."

Trip Computer Flow:

User types /trip → Assistant invokes skill
→ Skill executes: npx tsx src/index.ts --trip-computer
→ Same parsing + deduplication + analytics computation
→ Outputs dashboard to stdout
→ Assistant copies output to response (per skill instructions)

Session Caching:

5-second TTL cache in ~/.claude/session-stats/<session_id>.json
Cache invalidated when transcript mtime changes
~10ms cache hits vs ~200ms misses (90% hit rate typical)

Transcript Deduplication Algorithm:

// Group by requestId + model, take MAX tokens per request
const dedupKey = `${requestId}|${modelId}`;
requestUsage.input = Math.max(requestUsage.input, usage.input_tokens);
// Then aggregate into per-model totals

Cross-Project Agent Discovery:

Extracts agent IDs from main transcript: /agent-[a-z0-9]+/g
Searches all ~/.claude/projects/*/agent-*.jsonl files
Includes agent usage in cost calculation (agents are billable)

Zero Dependencies: Uses only Node.js stdlib (no npm install needed)

Features

✨ New in v0.13.0 (Multi-Line Status)

✅ Multi-line status display - Session metrics + tool activity + agents + todos ✅ Git branch integration - Shows branch name with dirty indicator (🌿 main*) ✅ Session duration - Time since session started (⏱️ 31m) ✅ Tool activity tracking - Top 5 tools by frequency with accurate counts ✅ Agent status display - Running and completed agents with duration ✅ Todo progress - Current task and completion status ✅ Accurate tool counts - Aggregates ALL tools (improvement over claude-hud's last 20)

Session Analytics

✅ Session health scoring - 0-100 automated assessment with 5-star rating ✅ Model mix visibility - See which models used with cost breakdown ✅ Cost drivers breakdown - Visual analysis of input/output/cache patterns ✅ Efficiency metrics - Tool intensity, response verbosity, cache hit rate ✅ Prioritized recommendations - Top 3 actions ranked by impact ✅ Billing mode differentiation - Adapts for API (💳) or Subscription (📅) users

Technical

✅ Best-effort estimates - Transcript-based calculations typically within 5-15% of /cost ✅ Model-aware pricing - All versions: Opus 3/4/4.5, Sonnet 3.7/4/4.5, Haiku 3/3.5/4.5 ✅ Accurate cache pricing - Model-specific multipliers (including Haiku 3 exception) ✅ Token deduplication - Groups by requestId to avoid inflation ✅ Cross-project agent tracking - Finds agent transcripts across all projects ✅ Real-time updates - Status line refreshes automatically

Why This is Valuable

Speedometer vs. Odometer:

Trip Computer = Speedometer (real-time insights for decision making)
/cost command = Odometer (authoritative billing from Anthropic)
Both are valuable: Use trip computer for session optimization, /cost for billing verification

Immediate Decision Making:

"This is getting expensive, let me switch to Haiku"
"Cache efficiency is great, stay in this session"
"This task used 3.8M tokens - worth tracking"

Cost Awareness:

Track session efficiency in real-time
Understand which workflows are expensive
Learn to use appropriate models
Improve cost efficiency over time

Session vs Billing:

Trip Computer = Real-time session insights
/cost = Final billing verification
Both are valuable for complete awareness

Pricing Reference (2026)

Model	Input	Output	Cache Write (5m)	Cache Read
Opus 4.5	$5/MTok	$25/MTok	$6.25/MTok	$0.50/MTok
Opus 3/4/4.1	$15/MTok	$75/MTok	$18.75/MTok	$1.50/MTok
Sonnet 3.7/4/4.5	$3/MTok	$15/MTok	$3.75/MTok	$0.30/MTok
Haiku 4.5	$1/MTok	$5/MTok	$1.25/MTok	$0.10/MTok
Haiku 3.5	$0.80/MTok	$4/MTok	$1/MTok	$0.08/MTok
Haiku 3	$0.25/MTok	$1.25/MTok	$0.30/MTok*	$0.03/MTok*

*Haiku 3 uses different cache multipliers: 1.20x for writes, 0.12x for reads

Subscriptions: Pro ($20/mo), Max 5x ($100/mo), Max 20x ($200/mo)

Known Limitations

Context Tracking

Limitation: Context tracking is only available when running as a status line command. Direct execution via bash doesn't receive stdin from Claude Code.

Why: Claude Code only provides stdin data (context_window, model) to processes invoked as status line commands.

Workaround: Use /context command for accurate context information.

Stats Reset with `/clear` Command

Current Behavior: When you use /clear, stats reset to zero (new session, new transcript).

Why: Each session has its own transcript file for isolation and simplicity.

Desired Future: Cumulative stats across /clear within same Claude Code instance.

Workaround: Note costs before /clear if tracking total spending across resets.

Disclaimer

These are session-level estimates from transcript data, typically accurate within 5-15% of the /cost command. Differences occur due to:

Web search costs (may not appear in transcript usage)
Background operations
Timing variations (transcript lag)
API measurement methods

For official billing amounts, always use the /cost command. For subscription users, these show API-equivalent costs - your actual usage is included in your plan.

Technical Architecture

Language: TypeScript with ES2022 modules Runtime: Node.js 18+ (via tsx for direct TS execution) Dependencies: Zero npm packages (stdlib only) Entry Point: src/index.ts Modules: Transcript parser, analytics computer, renderers, cache manager, usage API client Data Flow: stdin → transcript parse → analytics compute → cache → render

Future Enhancements

Potential improvements for future versions:

Support

Installation issues? Check TROUBLESHOOTING.md

Questions? Review CLAUDE.md for technical details

Want to contribute? See CONTRIBUTING.md

Need help? Verify Node.js 18+ is installed: node --version

Ready to use? Add to ~/.claude/settings.json and restart Claude Code!

Name		Name	Last commit message	Last commit date
Latest commit History 38 Commits
.claude-plugin		.claude-plugin
.github/workflows		.github/workflows
src		src
tests		tests
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
TESTING.md		TESTING.md
TROUBLESHOOTING.md		TROUBLESHOOTING.md
VERSION		VERSION
install.bat		install.bat
install.sh		install.sh
package-lock.json		package-lock.json
package.json		package.json
tsconfig.json		tsconfig.json
vitest.config.ts		vitest.config.ts

Navigation Menu

Repository files navigation

Claude Trip Computer

Quick Setup

Prerequisites

Automated Installation (Recommended)

Manual Installation

Need Help?

What You Get

Status Line

Trip Computer

How It Works

Technical Implementation

Features

✨ New in v0.13.0 (Multi-Line Status)

Session Analytics

Technical

Why This is Valuable

Pricing Reference (2026)

Known Limitations

Context Tracking

Stats Reset with `/clear` Command

Disclaimer

Technical Architecture

Future Enhancements

Support

About

Releases

Packages

Contributors

Languages

Folders and files

Latest commit

History

Repository files navigation

Claude Trip Computer

Quick Setup

Prerequisites

Automated Installation (Recommended)

Manual Installation

Need Help?

What You Get

Status Line

Trip Computer

How It Works

Technical Implementation

Features

✨ New in v0.13.0 (Multi-Line Status)

Session Analytics

Technical

Why This is Valuable

Pricing Reference (2026)

Known Limitations

Context Tracking

Stats Reset with /clear Command

Disclaimer

Technical Architecture

Future Enhancements

Support

About

Topics

Resources

Contributing

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

Stats Reset with `/clear` Command