Free · Windows 10 & 11 · 4.6 MB installer

Every AI coding agent on your PC.
And how much quota you have left.

LiberCode finds Claude Code, Codex, Antigravity, Grok, Copilot, Kimi Code and the rest — in terminals, editor extensions and desktop apps — and shows each account, plan, live usage limits and the sessions that are working right now. From the tray, at a glance.

Installer: no admin rights, Start menu entry, starts with Windows. Portable: one file, run it from anywhere. No runtime. An account is optional: create one.

LiberCode window: overview of Claude Code, Codex, Antigravity, Grok and Copilot with usage bars
11first-party agents
5 ssession refresh
1 minusage refresh (up to 30)
8 mslongest UI pause during a full scan
~1%of one CPU core

Everything it does

One small window and a tray icon. No cloud, and an account only if you want one.

Finds every install

  • Native CLIs, global npm, pnpm and bun packages, Python tools from uv and pipx
  • Extensions in VS Code, Insiders, VSCodium, Cursor, Windsurf, Antigravity, Qoder and other VS Code–family editors
  • Desktop apps, Microsoft Store apps, winget packages, and copies bundled inside other apps, like the Claude app's Code tab
  • The version and folder of every install; click one to open it
  • Older extension copies an editor keeps until restart, hidden unless you ask

Real session status

  • Running means working right now, measured from each agent's CPU activity
  • Idle sessions are open but quiet, so a forgotten terminal tab never counts as running
  • Open apps and background servers are shown as exactly that
  • Conversations inside app servers (Codex in VS Code, Antigravity, Copilot, the Grok app) count while they work
  • Rechecked every 5 seconds

Accounts and plans

  • The signed-in email or name and the plan for each agent
  • Separate accounts when they differ, such as the Claude CLI and the Claude app, or the Grok CLI and the Grok app
  • One GitHub account even when the CLI, the Copilot app and an editor each hold its sign-in
  • A clear “Not signed in” with the command that fixes it

Live usage limits

  • 5-hour, weekly and monthly limits, per-model quotas, premium requests, credits and extra usage
  • The exact time each one resets, counting down
  • Bars turn amber from 70% and red from 90%
  • Keeps the last reading when a vendor can't be reached, and backs off when one rate-limits
  • Refreshes every 1, 2, 5, 10, 15 or 30 minutes

A tab for every provider

  • An overview of every agent with its busiest quota
  • One tab per provider with every account, limit and install
  • The tab strip scrolls: mouse wheel, touchpad, arrow buttons, a draggable bar, ←/→ and Ctrl+Tab
  • Smooth scrolling, Windows 11 Mica, light or dark with Windows, sharp on any display

Lives in the tray

  • Hover the icon for the busiest quotas
  • Right-click for every agent and its headline number
  • Point at an agent to see its usage bars, account and installs, without opening a window
  • Refresh now, Settings and Show in window, one click away

Settings that stay out of the way

  • How often usage refreshes
  • Start with Windows, and start hidden in the tray
  • Keep running in the tray when the window closes
  • Show or hide older extension copies

Native and light

  • Win32 and Direct2D — no Electron, no .NET, no browser engine
  • 4.6 MB installer, 7.6 MB portable app, about 50 MB of memory
  • All network and file work runs in the background, so the window never freezes
  • Per-user installer with no admin prompt, a Start menu entry and a clean uninstaller
A provider tab showing Claude Code's limits and every install
Each provider's tab: limits, resets, and every install with what it's doing.
Tray menu listing every agent, with Claude Code's usage submenu open
Point at an agent in the tray menu to see its limits.
Settings: account, refresh interval, start with Windows, tray options
Refresh from every minute to every 30 minutes.

What each agent supports

First-party agents from the companies that make the models. How much the app can show depends on what each vendor exposes.

Full Partial Not yet
Install detection
11 full
Session status
11 full
Plan
10 full · 1 not yet
Account
7 full · 4 partial
Reset times
6 full · 1 partial · 4 not yet
Live usage
3 full · 6 partial · 2 not yet
AgentDetectionSessionsAccountPlanUsageResetsWhere usage comes from
Claude CodeAnthropicFullFullFullFullFullFullAnthropic usage API, live
CodexOpenAIFullFullFullFullFullFullChatGPT usage API, live
GitHub CopilotGitHubFullFullFullFullFullFullGitHub Copilot quota API, live
AntigravityGoogleFullFullFullFullPartialFullAntigravity's local server, live while the IDE is open
Kimi CodeMoonshot AIFullFullPartialFullPartialFullKimi usage API, live while Kimi Code keeps its sign-in fresh
GrokxAIFullFullFullFullPartialFullGrok CLI log, updated when the CLI starts
Muse CodeMetaFullFullFullFullPartialPartialMuse session log, only when the quota is used up
Qwen CodeAlibabaFullFullPartialFullPartialNot yetLocal token log
MiMo CodeXiaomiFullFullPartialFullPartialNot yetLocal session history
QoderAlibabaFullFullFullNot yetNot yetNot yetAccount only, from the app
Mistral VibeMistral AIFullFullPartialFullNot yetNot yetPlan only; Mistral has no Vibe usage endpoint

Partial account means the sign-in carries no email. Partial usage means live only while the agent runs, or local history instead of remaining quota.

How it works

  1. Detect. It looks where each agent installs itself: PATH, global npm and Python tools, every VS Code–family editor's extensions, app folders and Store packages.
  2. Read the sign-in. Each agent's own local sign-in, or the Windows credential store, says which account and plan you use.
  3. Ask the vendor. It asks that vendor's own usage endpoint, or reads the agent's local usage log when there is none.
  4. Watch the sessions. Every 5 seconds it checks which agent processes are working, which are idle, and which are only open apps or background services.

Your sign-ins stay where they are

Read-only. Sign-ins are never refreshed, rewritten or stored. Refreshing would sign your agents out.

Vendor only. A token goes only to the company that issued it, and only to read usage.

No telemetry. No analytics and no tracking. Settings and last readings stay in %APPDATA%\LiberCode. The optional account sends our server only its own email, name and password.

Questions

Windows says “Windows protected your PC”.

The app isn't code-signed yet. Click More info → Run anyway.

Do I need an account?

No. Every feature works without one. If you create one, sign in from the person icon at the top right of the window. Accounts don't unlock anything else yet.

Installer or portable?

The installer adds a Start menu entry, can start the app with Windows and adds an uninstaller in Settings → Apps, with no administrator rights. The portable .exe is the same app in one file: run it from anywhere and delete it to remove it.

Why doesn't it show an agent I use?

It lists current first-party agents. Retired products, such as Gemini CLI (replaced by Antigravity), and third-party wrappers aren't listed.

Does it work on macOS or Linux?

The detection and usage engine already runs on both as a terminal report. Native macOS and Linux windows are next.

Does it slow my PC down?

No. It uses about 1% of one CPU core while its window is open and almost nothing in the tray. Usage checks run in the background, so the window stays responsive.

See all your agents in one place

Version 0.1.0 · installer 4.6 MB, portable 7.6 MB · Windows 10 and 11 (x64)