What is MCP?
The Model Context Protocol (MCP) is an open standard for connecting AI assistants to external tools and data sources. Think of it as a universal adapter that lets AI models like Claude interact with your applications in a secure, structured way. Instead of manually copying data between your feed reader and your AI assistant, MCP enables direct, programmatic access.
Lion Reader’s MCP server is built with the official MCP TypeScript SDK and uses stdio transport for secure local communication. This means the connection stays entirely on your machine — no data is sent to external servers beyond what your AI assistant already does.
Available Tools
The Lion Reader MCP server exposes a comprehensive set of tools that mirror the web UI’s capabilities:
- list_entries — List feed entries with filters, search, and pagination
- get_entry — Get a single entry with full content
- mark_entries_read — Mark entries as read or unread in bulk
- star_entries — Star or unstar entries
- set_entry_score — Rate entries on a −2 to +2 scale
- count_entries — Get entry counts with filters
- save_article — Save a URL for later reading
- delete_saved_article — Remove a saved article
- upload_article — Upload Markdown content as an article
- list_subscriptions — List and search active subscriptions
- get_subscription — Get subscription details
Consistent Behavior
The MCP server uses the same services layer as the web UI, ensuring behavior is identical across interfaces. Whether you’re reading entries through your browser or asking Claude to summarize them, you’re accessing the same underlying data with the same permissions and filters.
Security and Usage
Access is secured via API tokens. Generate a token with the mcp scope from your account settings. The server is compatible with Claude Desktop and other MCP-supporting assistants. To run the server locally, use pnpm mcp:serve and configure your AI assistant to connect to it.