MCP Server

FAQs

Common questions about the Masset MCP Server

How does authentication work?

The Masset MCP Server uses OAuth 2.0 with JWT Bearer tokens. When you add Masset as a connector in your AI client, the client handles the OAuth flow automatically. Each user connects their individual Masset account through the AI client’s Connectors settings, which grants the client a token scoped to that user’s permissions.

This means the MCP server respects the same access controls you have in Masset — you will only see assets and data you have permission to access.

What are the available scopes?

The Masset MCP Server uses the following OAuth scopes to control access:

ScopeDescription
mcpBase scope required for all MCP access. Grants access to search context (taxonomy).
mcp:asset_readAllows searching for assets, retrieving metadata, and fetching previews.
mcp:asset_content_readAllows retrieving extracted text content (raw text, summaries, OCR, transcriptions).

The asset_analytics_search tool additionally requires analytics permissions within your Masset account.

What is the rate limit?

The MCP server allows 60 requests per minute per authenticated user. Normal conversational usage is well within this limit. If you encounter rate limit errors (HTTP 429), wait briefly before continuing your conversation.

Which AI clients are supported?

The Masset MCP Server is compatible with any MCP-compatible client that supports Streamable HTTP transport. Currently documented clients include:

  • Claude Web & Desktop (including Claude Cowork)
  • Claude Code
  • ChatGPT

If your AI client supports remote MCP servers via Streamable HTTP, it should work with the Masset MCP Server using the same endpoint and OAuth configuration.

What Masset data is accessible?

Through the MCP server, AI clients can access:

  • Search results (asset names, descriptions, and rankings)
  • Asset metadata (file details, tags, colors, detected objects, collections)
  • Extracted text content (raw text, summaries, OCR text, transcriptions)
  • JPEG preview images
  • Analytics data (download, view, and share counts)
  • Taxonomy information (categories, tags, collections)

All data access is governed by the connected user’s Masset permissions. The MCP server never exposes data that the user would not be able to access directly in Masset.

Is my data used for AI training?

No. Masset does not use your data for AI training. The MCP server simply provides your AI client access to your Masset content — it does not store, process, or use that data for any training purposes.

Data handling by the AI client (Claude, ChatGPT, etc.) is governed by that client’s own data policies and terms of service. Refer to your AI provider’s documentation for details on how they handle data.

Does the MCP server use any sub-processors?

The MCP Server does not currently use or require any additional sub-processors beyond the default set used by the Masset platform. If Masset introduces functionality requiring additional sub-processors in the future, these features will be explicitly opt-in for all customers.

Where can I submit feedback?

The Masset MCP Server is currently generally available and we welcome your feedback and suggestions. Your feedback is invaluable to help us expand the MCP server’s capabilities.

To submit feedback, bug reports, or feature suggestions, email us at support@getmasset.com.