lazyagent

One lazy eye on all your coding agents.

9 agents supported · TUI + desktop app + HTTP API · No server, no lock-in · MIT

A terminal UI, a macOS desktop app, and an HTTP API for monitoring all your coding agents from a single place. Watch sessions from Claude Code, Cursor, Codex, Grok CLI, Kilo, Kimi Code CLI, Amp, pi, and OpenCode — lazyagent doesn't replace your workflow, it watches it.

Official Website

lazyagent is transitioning to a full desktop application. The macOS build now ships as Lazyagent.app (brew install --cask illegalstudio/tap/lazyagent), which bundles the familiar lazyagent command alongside the GUI. Only interested in the CLI? Install the standalone build with brew install illegalstudio/tap/lazyagent-cli — same lazyagent command, TUI and HTTP API included, no GUI.

Inspired by lazygit, lazyworktree, and pixel-agents.

Support the project

⭐ If lazyagent is useful to you, consider starring the repo — it helps others discover it!

💛 Loving it? Consider becoming a sponsor to keep the project alive and growing.

lazyagent for iOS

Want to keep an eye on your agents from your pocket? lazyagent is available on the App Store for iPhone and iPad.

The iOS app is a paid app — and that's on purpose. Buying it is one of the easiest ways to support the project and keep development going. Thank you! 💛

That said, lazyagent and its API are fully open source. If you'd rather not pay for the app, you're more than welcome to build your own client on top of the API — that's exactly what it's there for. No hard feelings, the choice is yours. 🙂

News

📢 Session tools are here! Commands to find, reopen, search, and maintain your agent sessions — plus keep rate-limit usage visible:

• lazyagent prune — delete chat files older than N days or whose project folder no longer exists. Interactive agent picker, dry-run previews, and per-project row selection at the confirmation prompt.
• lazyagent compact — shrink session files in place by truncating bulky tool outputs, thinking blocks, and embedded images — sessions stay resumable with the originating agent. Supports Claude Code, pi, Codex, Grok, and Kimi.
• lazyagent search — search transcript-file agents (Claude, Codex, pi, Amp, Grok, Kimi) with highlighted snippets and an incremental local index.
• lazyagent limits — on-demand rate-limit / billing summary for Claude Code (5h + 7d), Codex (5h + 7d), Grok (monthly), Kimi Code, and Cursor (monthly, Models + API pools), with a detailed pace view available via --detailed.
• lazyagent sessions — list every session recorded for the current directory — across all agents — and reopen one with the originating agent's CLI. Interactive picker, --json for scripts.
• Outbound webhooks on session state transitions — send a signed JSON payload to Slack, a custom dashboard, or a CI endpoint whenever a session goes idle, waits for input, or changes state. See Webhooks.

Typical savings on a year of daily use: 80+ MiB reclaimed across the cleanup commands, with every rewrite validated and backed up by default.

Why lazyagent?

Unlike other tools, lazyagent doesn't replace your workflow — it watches it. Launch agents wherever you want (terminal, IDE, desktop app), lazyagent just observes. No lock-in, no server, no account required.

Terminal UI

macOS Desktop App

Detach the panel and lazyagent becomes a full desktop app — Dock icon, Cmd-Tab, native menus — with a card-grid dashboard (compact | rich | live density switch), per-card Resume/Editor actions with a right-click menu, and a Settings panel (terminal choice, editor, agents, API passphrase). Attach again to return it to the menu bar.

HTTP API

Install

Packaging change: the macOS cask now installs Lazyagent.app instead of a bare binary; the lazyagent command still works — the cask links it into Homebrew's bin at install time, and the lazyagent-cli formula provides the same command for CLI-only setups. The cask and the formula conflict on purpose: the app already includes the CLI.

Homebrew

Desktop app (macOS, universal binary — TUI + GUI + HTTP API):

brew install --cask illegalstudio/tap/lazyagent

Installs Lazyagent.app and links the lazyagent command into Homebrew's bin — the CLI works immediately, no first launch required.

CLI only (macOS, Linux — TUI + HTTP API, no GUI):

brew install illegalstudio/tap/lazyagent-cli

Go (TUI only)

go install github.com/illegalstudio/lazyagent@latest

Build from source

git clone https://github.com/illegalstudio/lazyagent
cd lazyagent

# TUI only (no Wails/Node.js needed)
make tui

# Full build with macOS app (requires Node.js for frontend)
make install   # npm install (first time only)
make build

Launch

lazyagent                    Launch the terminal UI (monitors all agents)
lazyagent --agent claude     Monitor only Claude Code sessions
lazyagent --agent grok       Monitor only Grok CLI sessions
lazyagent --agent kimi       Monitor only Kimi Code CLI sessions
lazyagent --api              Start the HTTP API (Bearer-token protected)
lazyagent --gui              Launch the desktop app (menu bar)
lazyagent --tui --gui --api  Run everything together
lazyagent prune --days N     Delete chat sessions older than N days
lazyagent compact            Shrink chat files by truncating bulky payloads
lazyagent search "query"     Search chat transcripts with snippets
lazyagent sessions           List and reopen sessions for the current directory
lazyagent limits             Show 5h / weekly / monthly usage summary
lazyagent passphrase         Set or rotate the HTTP API passphrase
lazyagent --help             Show full help

Documentation

Full documentation — supported agents, activity states, keybindings, configuration, the HTTP API, maintenance commands, and architecture — lives at:

• lazyagent.dev/docs — rendered website
• docs/ — Markdown sources in this repository, organized by topic:
  • Getting started — install, quickstart
  • Concepts — how it works, supported agents, activity states, session info
  • Interfaces — terminal UI, macOS GUI, HTTP API
  • Usage — CLI reference, directory-scoped sessions, recipes
  • Maintenance — prune, compact, search, and limits commands
  • Reference — configuration, architecture, development, roadmap

License

MIT