Overview

The Model Context Protocol (MCP) allows LLMs to interface with applications seamlessly. Ragie natively provides streamable-http MCP servers for each partition.

Each partition's MCP Server exposes the Retrieve tool, which performs a retrieval scoped to the given partition. This means any LLM or MCP-compatible application can directly query a partition and get back the chunks of data it needs, without additional setup.

Because every partition runs as its own MCP Server, users can connect only the data sources they want an AI to access. This provides a clear boundary around data and makes it easy to control context on a per-partition basis.

MCP support also standardizes how Ragie integrates with external AI applications. Instead of building and maintaining one-off connectors, the same MCP servers can be used across different clients (e.g. Claude Desktop, IDE plugins, or other MCP-enabled tools).

MCP is enabled for all partitions. For configuration settings, log in to Ragie.