Claude Context MCP Server

Finding an authentication routine or a related implementation starts with a natural-language query rather than an exact keyword match. Claude Context exposes this workflow through MCP for Claude Code and other coding agents. The index_codebase tool prepares a directory for retrieval using a configurable code splitter, while search_code queries that index for relevant code. Both take an absolute codebase path. Retrieved context can support investigating bugs, reviewing existing patterns, locating refactoring targets, or understanding the architecture before adding a feature.

Index maintenance is part of the same workflow: get_indexing_status reports indexing progress or completion, and clear_index removes the search index for an absolute path. When a search targets an unindexed codebase, the tool reports that indexing is needed. If indexing encounters a conflict on an already indexed path, its usage guidance calls for user confirmation before forced replacement. Code chunking uses AST analysis; incremental indexing uses Merkle trees to re-index only changed files. The project integrates with Zilliz Cloud for vector search and lets users configure embedding models, supported file extensions, and ignore patterns.