Connect Your AI Assistant to LabelGrid (MCP)

Updated July 2026.

LabelGrid publishes an official MCP (Model Context Protocol) server so you can connect your own AI assistant to your LabelGrid account and work in plain language. Instead of writing REST calls, you can ask your AI to draft a release, validate it, check its review status, or pull your analytics, and it uses the LabelGrid API under the hood.

The server is the npm package @labelgrid/mcp and is open source under the MIT license at github.com/labelgrid/labelgrid-mcp. It works with Claude Desktop, Claude Code, Cursor, or any MCP-compatible client.

What you can do

Once connected, you can ask your AI assistant to:

• Create and update draft releases and tracks — build out metadata conversationally
• Upload audio, artwork, and license documents to your releases and tracks
• Validate releases before submission so problems surface early
• Check review issues, quality reports, and delivery status
• Pull streaming analytics, royalty statements, and breakdowns
• Manage webhooks for asynchronous notifications
• Submit releases for distribution or take them down — available only when you explicitly enable these actions (see Safety and permissions)

Under the hood, the server exposes more than 80 tools across ten toolsets: identity, reference data, catalog (labels, artists, writers, and publishers), releases and tracks, review and quality, analytics, accounting and royalties, delivery status, webhooks, and distribution. File uploads are part of the catalog and distribution toolsets.

---

What you need

1. A LabelGrid account with API access. API access is part of LabelGrid’s API plans — see the API Overview for how to get it.
2. An API token. Generate one in your LabelGrid dashboard under Profile → API Tokens.
3. Node.js 20 or newer on the computer running your AI client.

---

Install

Add LabelGrid to your AI client’s MCP configuration:

{
  "mcpServers": {
    "labelgrid": {
      "command": "npx",
      "args": ["-y", "@labelgrid/mcp"],
      "env": {
        "LABELGRID_API_TOKEN": "your-api-token"
      }
    }
  }
}

Replace your-api-token with the token from your dashboard, then restart your client.

For Claude Code, it is a single command:

claude mcp add labelgrid -e LABELGRID_API_TOKEN=your-api-token -- npx -y @labelgrid/mcp

---

Guided setup

If you start the server without a token, it does not fail — it launches a guided setup mode instead. Add LabelGrid to your client without the LABELGRID_API_TOKEN value, then ask your AI to “set up LabelGrid” and it walks you through creating a token and where to put it. Your token itself is never pasted into the chat.

---

Safety and permissions

The server uses three permission tiers so you decide how much your AI can do:

• Reads are always available. Your AI can look up releases, tracks, reviews, analytics, statements, and delivery status at any time.
• Safe writes are on by default. Creating and editing drafts, updating metadata, and managing webhooks are enabled out of the box. You can turn them off with LABELGRID_ENABLE_WRITES=false, or force everything to read-only with LABELGRID_READ_ONLY=true.
• Consequential actions are off by default. Submitting a release for distribution, taking a release down, and immutable file uploads require LABELGRID_ENABLE_FULL_WRITES=true plus an explicit acknowledgment sentence, which is documented in the project README.

All of your account protections — validation, plan limits, and rate limits — are enforced by LabelGrid’s servers. The AI cannot bypass them, whatever permissions you grant it.

---

Limit which tools are available

If you only want to expose part of your account, set LABELGRID_TOOLSETS to limit which tool areas the server offers. This keeps the AI focused on the areas you choose.

---

Troubleshooting

• Token invalid or rejected — recreate the token in your dashboard under Profile → API Tokens and update it in your MCP configuration.
• You do not see an “API Tokens” page — your account does not have API access yet. See the API Overview for how to enable it.
• Not sure the server connected — the server prints its status when it starts, so check your client’s MCP logs.
• You want to disconnect — you can revoke a token at any time from your dashboard, which immediately stops all access.

---

FAQ

”Does LabelGrid have an MCP server?”

Yes. LabelGrid publishes an official MCP server as the npm package @labelgrid/mcp, open source under the MIT license at github.com/labelgrid/labelgrid-mcp. It connects Claude Desktop, Claude Code, Cursor, or any MCP-compatible AI client to your LabelGrid account so you can work in natural language.

”What can I do with it?”

Create and update draft releases and tracks; upload audio, artwork, and license documents; validate releases before submission; check review issues, quality reports, and delivery status; pull streaming analytics, royalty statements, and breakdowns; and manage webhooks. Submitting releases for distribution and taking them down are available only when you explicitly enable them.

”Is it safe to let an AI manage my account?”

The server has three permission tiers. Reads are always available; safe writes (drafts, metadata, webhooks) are on by default and can be turned off; and consequential actions (distribution submission, takedowns, immutable file uploads) are off by default and require you to enable full writes plus an explicit acknowledgment. All account protections — validation, plan limits, rate limits — are enforced by LabelGrid’s servers and cannot be bypassed by the AI.

”Do I need an API plan?”

Yes. The MCP server connects through the LabelGrid public API, so your account needs API access, which is part of LabelGrid’s API plans. See the API Overview for how to get access.

”Which AI apps are supported?”

Any MCP-compatible client, including Claude Desktop, Claude Code, and Cursor.

---

Related

• API Overview and Quickstart — how to get API access and authenticate
• Webhooks — asynchronous event notifications
• @labelgrid/mcp on npm
• labelgrid-mcp on GitHub

---

Need help?

If you have questions about the MCP server, contact our support team.