Setup Guide
Three steps from your first command to a growing garden. Takes about a minute.
Step 01 — Install
Connect your coding machine
Get your installation command
Sign in to VibeMon, choose Connect a device, then copy the complete command for Windows or macOS / Linux. Your one-time connection token is filled in for you.
Get installation command →- The command installs AttemptDB, connects your account, configures detected agents and starts background syncing.
- Keep the terminal open until installation finishes. On Windows, use PowerShell. On macOS, use your desktop session. Linux supports background sync without systemd; in session mode, rerun the installer after the environment restarts and keep its data directory.
- Existing local settings are preserved. Older VibeMon hooks are removed only after the first successful upload.
- After installing, run an agent task and check that new activity arrives. Pairing or an installation test alone does not confirm ongoing collection.
Step 02 — Code
Start coding. Your garden gathers energy.
Once installed, the work your agents do turns into energy for the garden automatically. Works with Claude Code, Gemini CLI, Cursor, and Codex CLI.
Edits, commits and passing tests become energy
Commits and passing tests count far more than plain edits. Agent hooks pick this activity up on your machine as it happens.
Claude Code hooks (9)
Step 03 — Grow
Open the app and watch your garden grow.
A sprout-headed jelly slime and its gardener friends plant, water, prune and harvest on their own — you just pick a focus. Make it yours in the app with slime colours, hats, outfits, eye styles and decor.
- Plant, water and harvest happen on their own
- Every day of AI coding is planted as a tree in your record forest
- Resting never hurts: nothing wilts, nothing punishes you
What We Collect
What leaves your machine is metadata about the work plus the conversation — your prompts and the agent's replies, secrets redacted — never commands, code, or tool output. The install script and client are fully open source — you can read every line at /install.sh.
- The conversation — your prompts and the agent's replies, with secrets redacted
- Lines of code — how many lines were added or removed (just the count)
- File path — e.g.
src/app/login/page.tsx(full path, not contents) - Tool type — write / edit / notebookedit / bash
- Project & git context — repo (e.g.
user/repo), current branch, and commit SHA - Working directory — the local folder path your agent is running in
- Commit message title — first line only, 200-char cap, on by default (toggle above)
- Session ID & local time — groups events from one session; hour/day/timezone for activity patterns
- Timestamps — when activity happened
- Commands you run, or their output
- Source code, diffs, or file contents
- Tool output of any kind
- Variable names, function bodies, or comments
- Environment variables or secrets (secrets in the conversation are redacted)
How it works — data flow
- 1Claude Code triggers a hook on each file edit
- 2The
attemptclient on your machine works out line counts locally and redacts secrets, leaving code, commands and tool output behind - 3Metadata about the work plus the conversation reaches the server — never commands, code, or tool output
- 4The server records the activity as garden energy
Troubleshooting
Verify hooks are installed
Reinstall
Get your installation command
Sign in to VibeMon, choose Connect a device, then copy the complete command for Windows or macOS / Linux. Your one-time connection token is filled in for you.
Get installation command →Command expired or key rejected?
Open Devices and generate a new command. Copy the complete command unchanged. It contains a one-time VibeMon token; you do not need a key from your AI provider.