--- title: "Introduction" description: "What antiburn is, how it works, and where to start." url: "https://antiburn.ai/docs/" --- ## What antiburn is antiburn is a local, open-source menu-bar app that watches the coding-agent sessions already on your machine and shows you where they're burning tokens, cost, and quota. It runs quietly in your menu bar (macOS) or system tray (Windows/Linux) - no Dock icon, no window at launch - and opens a popover when you click it. There's no CLI, no subcommands, and no config file to hand-edit. Everything is driven from the tray popover, the Settings window, a short first-run setup, and (on macOS) a small floating HUD. ## What it does antiburn discovers the session logs your coding agents have already written to disk, analyzes them on your machine, and turns that into three things: - **Activity and session detail** - what happened in each session, its token and cost breakdown, and which tools, models, and skills it used. - **Cost and quota visibility** - API-equivalent cost estimates for what you've spent, and, where available, your live plan usage straight from the provider. - **Findings** - a fixed set of checks that flag specific token-burning patterns, like sessions that got too deep or subagents running on an unnecessarily expensive model. See [How antiburn reads sessions](/docs/concepts/how-antiburn-reads-sessions/), [Tokens, cost and quota](/docs/concepts/tokens-cost-and-quota/), and [Hygiene score and findings](/docs/concepts/hygiene-score-and-findings/) for how each of these works. ## Local-first, by design Your session content stays on your device. antiburn needs no account, server, or backend of its own to do its main work. The two things it does contact are documented in full: a public model-price catalog from models.dev (no session data or credentials sent), and, in release builds, GitHub Releases for update checks. Anonymised product analytics are optional and can be turned off in Settings. ## Supported agents and platforms antiburn reads sessions from Claude Code, Codex, Cursor, GitHub Copilot, Cline, OpenCode, Kiro, Amp, Antigravity, Windsurf, and Pi. Coverage - which platforms, whether WSL is included, and how deep the analysis goes - varies by agent. See [Supported coding agents](/docs/agents/supported-agents/) for the full matrix. antiburn itself runs on macOS 13 or later (Apple silicon and Intel), Windows 11, and mainstream Linux desktops with a system tray. ## Where to start - [Installation](/docs/getting-started/installation/) - the install commands, what they verify, and where antiburn lands on each platform. - [First run and onboarding](/docs/getting-started/first-run/) - what setup asks for and why. - [Updating antiburn](/docs/getting-started/updating/) - how the in-app updater works. If something isn't working, [Common problems](/docs/troubleshooting/common-problems/) covers the usual causes.
Introduction
What antiburn is, how it works, and where to start.
What antiburn is
antiburn is a local, open-source menu-bar app that watches the coding-agent sessions already on your machine and shows you where they’re burning tokens, cost, and quota. It runs quietly in your menu bar (macOS) or system tray (Windows/Linux) - no Dock icon, no window at launch - and opens a popover when you click it.
There’s no CLI, no subcommands, and no config file to hand-edit. Everything is driven from the tray popover, the Settings window, a short first-run setup, and (on macOS) a small floating HUD.
What it does
antiburn discovers the session logs your coding agents have already written to disk, analyzes them on your machine, and turns that into three things:
- Activity and session detail - what happened in each session, its token and cost breakdown, and which tools, models, and skills it used.
- Cost and quota visibility - API-equivalent cost estimates for what you’ve spent, and, where available, your live plan usage straight from the provider.
- Findings - a fixed set of checks that flag specific token-burning patterns, like sessions that got too deep or subagents running on an unnecessarily expensive model.
See How antiburn reads sessions, Tokens, cost and quota, and Hygiene score and findings for how each of these works.
Local-first, by design
Your session content stays on your device. antiburn needs no account, server, or backend of its own to do its main work. The two things it does contact are documented in full: a public model-price catalog from models.dev (no session data or credentials sent), and, in release builds, GitHub Releases for update checks. Anonymised product analytics are optional and can be turned off in Settings.
Supported agents and platforms
antiburn reads sessions from Claude Code, Codex, Cursor, GitHub Copilot, Cline, OpenCode, Kiro, Amp, Antigravity, Windsurf, and Pi. Coverage - which platforms, whether WSL is included, and how deep the analysis goes - varies by agent. See Supported coding agents for the full matrix.
antiburn itself runs on macOS 13 or later (Apple silicon and Intel), Windows 11, and mainstream Linux desktops with a system tray.
Where to start
- Installation - the install commands, what they verify, and where antiburn lands on each platform.
- First run and onboarding - what setup asks for and why.
- Updating antiburn - how the in-app updater works.
If something isn’t working, Common problems covers the usual causes.