Iwe is a free, Rust-built MCP server tool from Iwe Org, available on the mcp platform. It transforms Markdown directories into a queryable knowledge graph for human users and AI agents. This tool functions as a memory system, connecting to AI tools via the Model Context Protocol. It offers database-style queries on local Markdown notes without requiring migration to a traditional database. Iwe supports installation via Homebrew or Cargo, or from source, providing a local-first approach to knowledge management.

What does Iwe connect to?

Iwe connects to AI tools such as Claude Desktop, Cursor, and Windsurf, functioning as an MCP server. It uses the Model Context Protocol (MCP) to link AI systems with data sources. Confirmed compatible AI clients include Claude Code and other tools speaking MCP. The Iwe MCP server explicitly supports the stdio transport protocol. Its documentation does not confirm HTTP, SSE, or streamable HTTP as supported transport protocols for the server itself, beyond the general nature of MCP.

Core features and tools

Iwe exposes several callable tools via its CLI and MCP server interfaces, including search, retrieve, create, and refactor. Core CLI commands are "find" for fuzzy and full-text search, "retrieve" to get content by key, and "tree" to display hierarchical structures. Additional commands include "iwe init --okf" to scaffold Open Knowledge Format bundles, "iwe schema validate" to check schema conformance, and "iwe find --filter '{type: …}'" to query OKF frontmatter. It also includes "iwe normalize -k <key>". AI agents gain memory skills like "/iwe:distill", "/iwe:init", and "/iwe:reflect".

Who is it for?

The primary audience for Iwe includes individuals and AI agents needing database-style queries on their Markdown notes without migrating them. It serves those who want structured access to Markdown notes for AI agents, enabling context retrieval and knowledge graph navigation. Coding agents benefit from a memory system that stores session information as Markdown files within a project's Iwe workspace. This ensures memory persists with the project, viewable as Git diffs. Skip this if you require a traditional database for your structured data.

How to get started

No credentials are required. The system operates on explicit user permission for each action with AI clients like Claude Desktop, relying on a local-first approach. The MCP server supports the stdio transport protocol.

Use cases

• Structured access for AI agents: AI agents can use the CLI and MCP server for structured access to Markdown notes, retrieving context and navigating the knowledge graph.
• Memory for coding agents: The Iwe plugin for Claude Code allows coding agents to store session information directly in the project's Iwe workspace as standard Markdown files.
• IDE-like features for Markdown notes: Iwe integrates with editors supporting the Language Server Protocol (LSP), like VS Code, Neovim, Zed, and Helix, to provide search, refactoring, and autocomplete.

Iwe offers a unique local-first solution for managing Markdown-based knowledge, acting as an MCP server for AI agents and human users. Its Rust foundation and Git-based versioning support a non-cloud approach. The tool extends Markdown capabilities with database-like queries and IDE features, making local notes highly functional. Try preparing context for an AI conversation using "iwe find" and "iwe retrieve".