Analytics for AI-assisted development

Measure your vibe.

The observability layer Claude Code is missing.

Project-level usage, session history, tool patterns, and long-term trends for your AI coding workflow. See where usage comes from, how it changes over time, and which projects drive it.

$curl -fsSL https://vibenalytics.dev/install.sh | bash

macOS and Linux. Free with an open source CLI - inspect exactly what gets collected before you install it.

WHY VIBENALYTICS

Stop guessing. Start measuring.

Claude Code gives you output. Vibenalytics gives you context. Pick the blind spot you want to fix first.

Problem

Your usage is hidden from you

Claude Code subscription plans don't show token counts in project context. You can't tell which session or project actually drove the usage.

Solution

See the API-equivalent value per project

Input, output, cache read, and cache creation tokens are tracked separately per request. Cost becomes a reference point for attribution, not just a raw number.

vibenalytics
This month
$842
Tokens42.1M
Prompts847
Sessions84
Lines by language
TypeScript
+1,842-467
Vue
+924-281
Rust
+612-198
CSS
+341-89
backend-api
$12.8028.6M tok
marketing
$5.4012.3M tok
cli-tool
$3.608.1M tok
GET STARTED

Up and running in 60 seconds

1

Install the CLI

One command. Downloads the binary, adds it to your PATH, and sets up Claude Code hooks automatically.

$curl -fsSL https://vibenalytics.dev/install.sh | bash
2

Launch and log in

The TUI guides you through account creation, project selection, and optional history import - all in one flow.

$vibenalytics
3

Done. Start coding.

Once enabled, usage metadata is captured in the background and synced to your dashboard automatically.

$vibenalytics status

Run this anytime to verify everything is working.

INTEGRATIONS

Built for Claude Code
with more clients coming

Claude Code is supported today. Other AI coding tools are planned, but the current product is focused on doing one workflow well.

Claude Code
Soon
Gemini CLI
Soon
OpenAI Codex
Soon
OpenCode

Missing your tool? Let us know

WHAT WE TRACK

Metadata with context

Usage metadata is aggregated across sessions, projects, and time ranges so you can understand patterns without exposing prompts or code.

Captured metadata

tokensnumber
linesChangednumber
numLinesAdded
toolCallsarray
skillsarray
subagentCountnumber
modelstring
contextTokensnumber
requestCountnumber
timestampdate

Tokens, lines changed per language, tool calls, skills, subagents, and timestamps - enough to understand behavior without collecting content.

Privacy-first

Only metrics, never content

Your prompts, code, and file contents are never captured or transmitted.

Paths hashed locally

Project paths are hashed before syncing. Raw filesystem paths and identifiers never leave your machine.

Open source CLI - zero trust required

The entire data collection layer is on GitHub. Read the code, verify what's captured, and decide for yourself.

Low-friction capture

Runs silently in the background

Hooks fire asynchronously. You never notice it's there.

Built to stay out of the way

Collection runs asynchronously so observability does not become workflow friction.

Automatic sync

Metrics flow to your dashboard without manual exports or transcript parsing.

RESEARCH

Understanding AI coding with real-world data

With your consent, anonymized and aggregated metrics help answer questions that synthetic benchmarks can't. Individual sessions are never shared.

Are models actually getting better?

Track tool success rates, token efficiency, and iteration counts across model versions. Real production data, not synthetic benchmarks.

How do coding patterns change over time?

Discover which workflows lead to faster completion. See how tool usage evolves as developers gain experience with AI coding.

What does real AI coding look like?

Aggregate anonymized data across thousands of sessions to understand how AI-assisted development actually works in the real world.

All contributions are opt-in, anonymized, and aggregated. Your code and prompts are never part of any dataset.

FAQ

Frequently asked questions

Stop guessing. Start measuring.

Install Vibenalytics in 30 seconds. The CLI is open source - see exactly what data is collected before you trust a single byte.

$curl -fsSL https://vibenalytics.dev/install.sh | bash