FAQ.md

TIDAL MCP FAQ

What is TIDAL MCP?

TIDAL MCP is a local Model Context Protocol server that lets compatible AI assistants search and organize a TIDAL account through typed tools. It covers catalog lookup, recommendations, lyrics, favorites, playlists, folders, mixes, discovery pages, and safety-gated account mutations.

Is this an official TIDAL product?

No. This is an independent, unofficial open-source project. It is not affiliated with, endorsed by, or sponsored by TIDAL. TIDAL is a trademark of its respective owner.

Is this the most complete TIDAL MCP server?

It is the broadest public implementation found in the dated September 4, 2026 comparison. TIDAL MCP now exposes 122 registered tools; the next-largest reviewed public implementation exposes 78. Tool count is not the only measure of quality, so the comparison also records schemas, mutation safety, tests, transports, and notable scope. Read the evidence and limitations in COMPETITIVE_MATRIX.md.

Which AI clients can use it?

Any MCP client that can launch a local stdio server should be compatible. The implementation has been exercised through a real MCP stdio subprocess and configured locally with Codex. Claude Desktop, Claude Code, and Cursor also accept stdio MCP configurations, although client UI and configuration paths can change independently.

Does it require TIDAL developer credentials?

No developer client ID or secret is required. Authentication uses TIDAL device login through the pinned unofficial tidalapi adapter. This offers broad consumer-account coverage but can require maintenance when TIDAL changes its private endpoints.

Where are my TIDAL credentials stored?

The OAuth session is stored outside the repository in the operating system's private application data directory. Authentication happens through the separate tidal-auth command. Tool responses serialize only allowlisted public fields and cannot expose the session object or OAuth tokens.

Never commit or share session.json.

Can it analyze an entire playlist?

Yes. The server can list the authenticated user's playlists, resolve an exact playlist, paginate through all its tracks or mixed items, compare reported counts, and return structured metadata for analysis. An AI assistant can calculate duration, artist and album concentration, eras, explicit content, duplicate candidates, and ordering patterns.

Musical attributes not exposed by the adapter—such as a guaranteed BPM, key, mood, or acoustic profile for every track—must not be presented as API facts.

Can it create or edit playlists?

Yes, when writes are explicitly enabled. Playlist creation, editing, deletion, clearing, merging, visibility, item insertion, removal, and reordering are covered. Favorites and playlist folders can also be managed.

How are accidental writes prevented?

Writes are disabled by default. When enabled, a named preview tool creates a private local draft containing the exact action, parameters, target metadata, expiry, and destructive flag. A separate commit tool requires the draft's single-use approval token. The assistant should call it only after the user approves the displayed preview.

Does it download music or bypass DRM?

No. Media downloading, manifest extraction, DRM handling, and file conversion are explicit non-goals. Some read tools return temporary, account-scoped playback metadata or URLs provided by the upstream adapter.

Does it control speakers or TIDAL Connect devices?

No. UPnP, DLNA, and device-network control are outside the current security boundary. They could be implemented as a separate optional component, but they are not counted as TIDAL API coverage.

Why are there 122 tools instead of one raw API tool?

Individual allowlisted tools are easier for an AI client to discover, validate, audit, and annotate correctly. An unrestricted raw endpoint would bypass input schemas, public-field serialization, and the mutation approval design.

Why can playlist metadata and fetched item counts differ?

TIDAL can return a cached count on a playlist object while its dedicated count endpoint and paginated item endpoint return a newer value. Agents should compare the dedicated track count, item count, and fetched pages before declaring the discrepancy an error.

Is Linux supported?

The project is designed for macOS and Linux. macOS has been tested against a real account. Linux is covered by deterministic CI but should still be treated as community-supported until more live account reports are available.

Is there a hosted remote server?

No. Version 1.0 is intentionally local-first and uses stdio, keeping account sessions and approval drafts on the user's machine. A future remote service would need a separate OAuth, tenancy, origin-validation, abuse-prevention, and privacy design.

How can I verify the claims?

Run the formatter, linter, full statement-and-branch coverage suite, real stdio smoke test, and package build documented in the main README. The tool inventory is defined in src/tidal_mcp/catalog.py and src/tidal_mcp/server.py.