Budi

Free

Local-first cost analytics for AI coding agents. Tracks token usage and spend across Claude Code and Cursor.

FreeFree tier

Type

Open Source

Visit Website GitHub (38 stars)

About Budi

Budi is a local-first, open-source cost analytics daemon and CLI for AI coding agents. It tracks token usage and spend across multiple agents including Claude Code, Cursor, Codex CLI, Copilot CLI, and Copilot Chat (VS Code/JetBrains), breaking down costs by repository, branch, ticket, and file. Data stays on the user's machine by default, with an optional cloud sync that sends only aggregated metrics (prompts and code never leave the machine). Features per-message cost attribution, session health detection (context bloat, cache degradation, cost acceleration), and a live status line for VS Code/Cursor. The Rust binary is approximately 13 MB, requires zero configuration, and supports one-time historical imports.

Key Features

Tracks multiple AI coding agents (Claude Code, Cursor, Codex CLI, Copilot CLI, Copilot Chat) in one place

Per-message cost and token attribution, not just session totals

Breakdown by repository, branch, ticket, and file

Session health detection: context bloat, cache degradation, cost acceleration with actionable tips

Live status line in Claude Code, Cursor, and VS Code showing rolling 1d/7d/30d spend

Local-first by default; optional cloud sync sends only aggregated metrics (prompts and code never leave machine)

Rust binary (~13 MB) with zero configuration and minimal footprint

One-time historical import capability via `budi db import`

Supports Homebrew (macOS/Linux), shell script, and Windows PowerShell installation

Ecosystem includes cloud dashboard, VS Code/Cursor status bar extension, JetBrains plugin

Pros & Cons

Pros

Local-first design ensures privacy of prompts and code
Free and open-source
Lightweight Rust binary with no dependencies
Supports multiple major AI coding agents
Granular per-message cost breakdown
Actionable session health insights
Optional cloud sync for team-wide dashboards

Cons

Limited to specific AI coding agents; some like Gemini CLI are deferred
JetBrains support is partial (session discovery + daily spend only)
Cloud sync is optional but may require additional setup for team use
Requires command-line installation and initial configuration
No native mobile or web app beyond the cloud dashboard

Best For

Monitor and analyze AI coding agent spending across multiple projectsIdentify which repositories, branches, or tickets are driving token usageDetect inefficient agent sessions (context bloat, cache degradation) and optimize usageTrack per-developer or per-team AI cost allocation in collaborative environmentsIntegrate cost visibility into daily coding workflow via IDE status line

FAQ

Which AI coding agents does Budi support?

Budi fully supports Claude Code, Codex CLI, Cursor, Copilot CLI, and Copilot Chat in VS Code/Cursor. JetBrains Copilot Chat has partial support (session discovery and daily spend). Gemini CLI support is deferred.

Is Budi really free?

Yes, Budi is open-source and free to use. The core daemon and CLI are available on GitHub under an open-source license.

How does Budi handle privacy?

By default all data stays on your machine. The optional cloud sync sends only aggregated daily metrics; prompts, code, and responses never leave your machine.