The beat library built for how producers actually work.
Not a spreadsheet. Not a DAW. Not a marketplace. BeatOS is a local-first home for every beat on your drive — catalog it once with the metadata that beats, let AI do the tagging grind, and . Runs as a native or right in your . Single-user, offline, no account, no telemetry.
Local-first beat library built for producers — catalog, BPM/key analysis, license tiers & PDFs, tagged-MP3 + loopkit export, and assisted multi-platform publishing. Chat with your catalog via a built-in AI Agent (Claude · ChatGPT · DeepSeek), or wire it to any MCP client. Desktop + web.
Sterne
30
Tools
—
Uptime
—
sells
export or publish to the platforms where you actually sell
I'm a beatmaker. Like most of us, my catalog lived in a spreadsheet, and every beat got its metadata typed out by hand into each platform I sell on. BeatOS is what I built instead — one canonical catalog, with three things nothing else puts together:
🎯 Purpose-built for selling beats — license tiers with multi-currency pricing, producer credits, tagged/untagged/loop/stems, BPM & key. Not a generic media manager bent into shape.
🤖 AI-native — talk to your catalog in plain language, draft genre/mood/tags from the cover, or wire it to Claude / Codex over MCP. Every edit waits for your approval.
🚀 Built to ship, not just store — export clean per-platform metadata and packaged loopkits / beat packs for free; Pro adds assisted publishing that drives a real browser and hands back at the platform's human-verification step.
A catalog made to sell. Every beat in a real SQLite database: title, BPM, key, genre + mood (multi-value), producer credits, tags, license tiers with multi-currency pricing, description, and per-role audio — tagged/untagged, loop, stems (WAV/MP3/FLAC) — plus cover art. Soft-delete trash with restore; rename or merge a producer across your whole catalog in one click. On a sale, generate a per-buyer license PDF (English or 中文) from any tier, and export a tagged MP3 with the title, producer, BPM, key, and cover baked into the file's ID3 tags.
Find any beat instantly. Search the whole catalog with stackable filters — BPM range, key, genre, mood, producer, has-audio — and recent searches one click away. The player queue follows whatever you're looking at.
Loopkits, beattapes & beat packs. Curate beats into a list, then package it to send out — a loopkit, a beattape, a singer's beat pack. Pick per-track and per-file what goes in (bulk-select all WAVs, all MP3s), and export as a ZIP or a plain folder, one subfolder per track.
Analysis & AI tagging. On-demand BPM + key detection with per-field confidence scores (Essentia, or the permissive librosa fallback) — one track or the whole library at once, auto-filling empty fields. Optional AI tag suggestions (bring your own key) draft genre/mood/tags from the cover + title for you to review field by field.
Chat with your catalog. A built-in AI Agent in the sidebar — ask in plain language ("find my untagged 2024 beats", "tag these as chill", "add a WAV tier") and the model actually reads and edits your catalog. Works with Claude, ChatGPT, or DeepSeek using your own key, stored on your device. Edits apply under your permission setting and are logged; destructive actions ask for a confirm first; read-only mode refuses writes. Conversations save locally and resume next launch.
Drive it from your own AI. A first-class MCP server exposes the library to Claude Code, Claude Desktop, Codex, and any MCP client — 23 tools free, 28 with Pro — so an agent in your terminal can organize, tag, and prep per-platform metadata for you.
Assisted publishing (Pro). Publish a beat without leaving BeatOS: the engine drives a real browser and pauses at the platform's verification step, with a Publish Center showing live per-platform session health. Live targets today: 抖音 (Douyin) promo-video and NetEase 激灵 (BeatSoul), with BeatStars on the roadmap. Pro build only — the free build greys it out.
And it feels like a product, not a database. Spotify-style cards and Coverflow, a floating player, an animated WebGL backdrop (Aurora or ASCII glyph-rain), a glowing search orb, and a tunable glass panel-opacity layer — all bundled to run fully offline. Fully bilingual (English / 中文), with an independent control for how genre/mood tags display.
Desktop or browser — one codebase, two front ends
BeatOS ships as a native desktop app (Electron) and a browser app (a local web SPA served by the same Python sidecar). One React codebase builds both, so they stay in lockstep — Electron-only powers (native file dialogs, reveal-in-Finder, drag-out) route through a thin platform seam with a same-origin web implementation behind it.
Desktop — the full native experience. make dev from source today; packaged installers land at v0.1.0.
Browser — make web builds the SPA and serves it at http://127.0.0.1:8765: same backend, same library, near-identical UI, no Electron build. The easy way to run on Windows or anywhere today.
Both are local-first and offline — the browser app talks only to 127.0.0.1. (Remote/LAN access and a mobile layout are planned.)
Two ways to put AI to work
BeatOS is built for the AI-agent era. Cataloging, tagging, and (soon) publishing are meant to scale without manual labor.
In-app — the AI Agent chat needs nothing but your own API key (Claude / ChatGPT / DeepSeek). Turn on AI Assist in Settings, paste a key, and start asking. Reads run directly; edits respect your permission mode and are logged.
Your own AI client — connect over MCP. Verified clients: Claude Code · Claude Desktop · Codex CLI/App · any client speaking stdio JSON-RPC.
You stay in control either way. Your MCP client gates each tool call (allow / deny per action), and every write the app makes is recorded in an Agent Actions dashboard — what changed, when, the result. Want a hard stop? Switch the agent to read-only and writes are refused. Batch tools (create_tracks ≤100, attach_assets ≤500) turn a 50-track import into one action instead of a hundred.
You: "Tag every beat above 140 BPM with no genre as 'Trap' or 'Drill',
from the cover and title."
Claude: list_tracks(bpm_min=140) → 12 tracks → drafts a patch
update_tracks(items=[...]) ← your client asks you to allow the call
You: Allow
Claude: applied — the 12 edits land and show up in Agent Actions
No server. The sidecar binds 127.0.0.1 and serves the browser front end same-origin. Nothing leaves the machine — including your conversations with the AI.
No account. Single-user. No login, no sync, no cloud.
No telemetry. Zero outbound calls from the app itself; your API key stays on your device.
Your files stay put. BeatOS references paths; nothing is moved or renamed unless you ask.
Your data is yours. One SQLite file in your per-user app-data folder (kept off cloud-synced directories where SQLite can corrupt; configurable in Settings). Open it with any tool.
Install & run
Packaged desktop installers arrive at v0.1.0. Until then, run from source. The browser front end needs no packaging.
Targets: macOS 12+ · Windows 10+ · any modern browser. (Linux: dev + web only.)
make sync && (cd apps/desktop && npm install) # one-time setup
make dev # desktop: Electron + sidecar
make web # browser: build the SPA + serve it, open a tab
No terminal? Double-click start-beatos.command (macOS) or start-beatos.bat (Windows) at the repo root — it checks/installs prerequisites, then launches the browser or desktop app.
The MCP server lives at packages/beatos-mcp. It bridges your MCP client (stdio) to the app's sidecar over local HTTP. It attaches even when BeatOS is closed (showing a beatos_status tool), and the full library tools appear automatically once you open the app — no client restart needed.
Install deps:uv sync from the repo root (creates .venv). Re-run after pulling.
Start BeatOS and leave it open.
One-click setup (recommended):Settings → AI Integration → click your client. BeatOS merges the config (with a .beatos.bak backup) or runs claude mcp add for you.
Manual fallback — register it yourself (--directory must be the absolute repo path):
Local-first beat library built for producers — catalog, BPM/key analysis, license tiers & PDFs, tagged-MP3 + loopkit export, and assisted multi-platform publishing. Chat with your catalog via a built-
Local-first beat library built for producers — catalog, BPM/key analysis, license tiers & PDFs, tagged-MP3 + loopkit export, and assisted multi-platform publishing. Chat with your catalog via a built-in AI Agent (Claude · ChatGPT · DeepSeek), or wire it to any MCP client. Desktop + web.
Tools
Tools
Noch keine Tool-Liste. Remote-Server erhalten eine nach dem ersten Health-Check.
FAQ
Fragen
Wie installiere ich Beatos?
Kopieren Sie auf dieser Seite die stdio-Konfiguration für Cursor, Claude Desktop, Claude Code oder VS Code. Sie nutzt die Installation des Herausgebers, keinen PlugMCP-Proxy.
In diesem Eintrag ist keine erforderliche Umgebungsvariable aufgeführt.
Welchen Transport verwendet Beatos?
Beatos läuft lokal über stdio.
Wann wurde Beatos zuletzt geprüft?
Zuletzt geprüft: 2026-11-05. Local-first beat library built for producers — catalog, BPM/key analysis, license tiers & PDFs, tagged-MP3 + loopkit export, and assisted multi-platform publishing. Chat with your catalog via a built-in AI Agent (Claude · ChatGPT · DeepSeek), or wire it to any MCP client. Desktop + web.