VibeDeck shows you what every AI coding tool on your machine is burning, in real time, all in one place.
Local-first. Multi-provider. Mac-native as the premier surface, with CLI and dashboard access for everyone. Branch-aware when local evidence supports it.
AI coding spend is scattered across local logs, provider folders, app state, and session files. VibeDeck turns that into one local view of what is active now, what it has burned, and where that usage belongs.
The product is built around three promises:
- See live spend across the AI coding tools you actually use.
- Keep the data local on your machine instead of sending prompts or transcripts to a hosted analytics service.
- Drill into projects and branches when attribution is available, without hiding existing non-git local work.
VibeDeck is not a hosted team telemetry product. If you want to share a snapshot, use exports, screenshots, or README banners.
VibeDeck is for developers who run AI coding tools locally and want one surface for current sessions, local usage history, provider status, and project context. The Mac app is the premier surface, with the CLI and browser dashboard still available when you want terminal workflows or a local web view.
For team sharing, VibeDeck stays intentionally local-first. Share exports, screenshots, or README banners instead of uploading developer telemetry to a hosted service.
brew install ivasuy/tap/vibedecknpm install -g vibedeck-cliRun without installing globally:
npx vibedeck-cli serveAdditional local adapters are present for Every Code, Goose, Crush, OMP, Pi, Droid, Qwen, Cursor Agent, IBM Bob, Roo Code, KiloCode, CodeBuddy, and Craft where local files or databases are available. Some providers are hook-based. Others are passive readers over local JSONL, SQLite, CSV, or native app state. Attribution depth depends on the local records each provider exposes, so VibeDeck is designed for mixed-runtime environments rather than single-provider lock-in.
- Live sessions and active workstreams
- Provider and model usage where local data is available
- Project and branch rollups for attributed local sessions
- Existing non-git local folders as visible projects
- Exact billable-token views where provider payloads expose billable token fields
- Per-day cost heatmaps for active-day and best-day widgets
- Historical local usage preserved in SQLite
- Local integration health, sync status, and doctor diagnostics
VibeDeck shows what is running now, not just what ran earlier. Live pages combine real-time session state with preserved historical local usage, so active projects do not lose their prior context when a session goes stale and comes back later.
When local records include enough context, usage can roll up under the engineering structure people work in: project, repo, worktree, branch, and session. Existing local folders without git metadata still remain visible instead of being discarded.
The packaged macOS app and desktop widgets let VibeDeck live outside the browser and act more like a daily operating surface than a hidden developer tool. The dashboard and CLI continue to cover local web and terminal workflows.
Default local state lives under ~/.vibedeck/, with canonical usage stored in SQLite and compatibility queue exports preserved alongside it. This keeps VibeDeck local-first while still giving you stable historical rollups and reconciliation surfaces.
Rebuilds use staged SQLite promotion, projection freshness, recent-first processing, dirty branch-fact materialization, and grouped-flush caches so new users can start from local provider files without blocking the UI on the full historical repair path.
VibeDeck also includes supporting surfaces for developers who want deeper local context after the core live-spend view is working.
VibeDeck keeps track of local integrations, skill-related runtime state, provider hooks, README sync, native install bootstrap, and local health surfaces through status, doctor, and setup flows. These are power-user operations for maintaining an AI-heavy local setup.
Dual-banner README surfaces provide optional profile-level visibility and project-level context.
Profile-ready usage surface with cross-provider model mix, token scale, estimated cost rollups, and trend context.
Repository-local usage surface focused on the active project path, with model split, token mix, estimated cost, and snapshot context.
Initialize local integrations:
vibedeck initSync local usage into the canonical database:
vibedeck syncStart the local dashboard:
vibedeck serveThen open:
http://127.0.0.1:7690
vibedeck serve
vibedeck sync
vibedeck status
vibedeck doctor
vibedeck diagnostics
vibedeck optimize --scan
vibedeck auth show
vibedeck auth rotate
vibedeck readme-sync status
vibedeck project-readme-syncMore examples live in docs/COMMANDS.md.
Default local state:
~/.vibedeck/
auth.token
cache/pricing.json
tracker/
vibedeck.sqlite3
cursors.json
queue.jsonl
project.queue.jsonl
diagnostics/
vibedeck.sqlite3 is the canonical local store for sessions, branches, projects, usage buckets, billable token rollups, projection freshness, and historical audit state.
Public dashboard routes today include /dashboard, /live, /branches, /settings, /skills, /widgets, /compare, /models, /yield, /export, /optimize, and /plan.
No. VibeDeck is local-first. It reads local usage signals, computes rollups locally, and stores state under ~/.vibedeck/.
No. VibeDeck is designed for mixed-runtime AI coding workflows and can ingest usage from multiple local providers into one product.
No. VibeDeck includes a Mac app, dashboard, CLI, widgets, local integrations, health/debug tooling, and optional power-user surfaces described above.
Provider billing pages usually stop at account-level usage. VibeDeck uses local provider records to connect usage to project, branch, worktree, folder, and session context when that attribution is available.
MIT