pyckle-mcp gives your AI editor persistent memory of your codebase. Index once with index_codebase, then search forever with natural language via search_code. Works with Claude Code, Cursor, Windsurf, VS Code, and Zed. Additionally, indexes Obsidian vaults, Notion workspaces, and git history. Session context, dependency graphs, and autonomous iteration loops included.
Overview
pyckle-mcp provides an MCP server with 18 tools focused on indexing and searching codebases, personal knowledge bases in Obsidian, and Notion workspaces. It enables programmatic access to build and query indexes of source code and notes, supporting efficient retrieval in development and research workflows.
Key Capabilities
- search_code: Queries indexed codebases for specific functions, classes, or snippets using semantic or keyword search.
- index_codebase: Scans and indexes Git repositories or local directories, creating searchable structures from source files.
- index_stats: Retrieves metrics on indexed data, such as file counts, languages, or index size.
- index_obsidian: Indexes Obsidian vaults, making Markdown notes and linked files searchable.
- index_notion: Indexes Notion pages, databases, and blocks for content retrieval.
These tools support incremental updates and handle large-scale data.
Use Cases
- A developer indexes a monorepo with index_codebase, then uses search_code to find implementations of an API across modules.
- A researcher indexes an Obsidian vault via index_obsidian and queries notes with search_code for research references.
- A team maintains Notion docs; index_notion followed by search_code locates specs during code reviews.
- Project leads run index_stats on codebases to track language distribution and growth.
Who This Is For
Software developers managing large codebases, technical writers organizing documentation in Obsidian or Notion, and teams needing unified search across code and notes. It suits solo contributors or small teams integrating search into IDEs or AI assistants.