LINZA
Give Claude Code or Cursor a local MCP sidecar that indexes your Markdown vault, surfaces review intents, and grows knowledge under your supervision instead of scattering notes in chat.
Overview
LINZA is an MCP server for the Build phase that indexes a local Markdown vault and exposes supervised artifact and knowledge-growth tools to your coding agent.
What is this MCP server?
- Stdio MCP server (linza-mcp) installable via uvx from PyPI at version 0.2.2
- LINZA_VAULT points indexing at a Markdown workspace (default ./vault)
- Embedding providers: lmstudio, openai, or ollama with configurable URL, model, and optional LINZA_EMBED_KEY
- Local-first sidecar for Markdown artifacts, review intents, and supervised knowledge growth
- Keeps agent retrieval on your disk rather than re-pasting long docs into every session
- Server version 0.2.2
- Package identifier linza-mcp on PyPI
- stdio transport
Community signal: 3 GitHub stars.
What problem does it solve?
Agents forget your written specs unless you copy-paste vault files every session, and unconstrained edits to notes are risky.
Who is it for?
Solo builders maintaining a Markdown or Obsidian-style vault who want semantic search and agent tools without sending notes to a third-party knowledge cloud.
Skip if: Teams that need real-time collaborative editing, CRM-style customer data, or a zero-setup cloud notebook with no local embedding stack.
What do I get? / Deliverables
After registration, your agent queries and proposes changes against a bounded LINZA vault with review-oriented intents instead of improvising from chat memory.
- Agent-callable tools over your indexed Markdown vault
- Review-intent oriented knowledge updates within configured bounds
- Embedding-backed retrieval when LINZA_EMBED_PROVIDER and model are configured
Recommended MCP Servers
Journey fit
Canonical shelf is Build because solo builders wire LINZA while setting up agent-accessible documentation and artifact workflows alongside the product. Docs is the primary facet: Markdown workspaces, vault indexing, and supervised knowledge updates are documentation and knowledge-base operations.
How it compares
Local Markdown knowledge MCP sidecar, not a hosted Notion connector or a general web-scraping browser automation server.
Common Questions / FAQ
Who is LINZA-MCP for?
Indie developers and technical writers who keep product and agent instructions in Markdown and use Claude Code, Cursor, or Codex with MCP.
When should I use LINZA-MCP?
Use it when you are building or maintaining a local docs vault and want agents to search artifacts, flag review intents, and grow knowledge under explicit supervision.
How do I add LINZA-MCP to my agent?
Install the PyPI package (uvx linza-mcp), set LINZA_VAULT and optional LINZA_EMBED_* variables, then add a stdio MCP entry pointing at that command in your host’s MCP config.