Lexomni MCP

Persistent memory & semantic search for AI agents using Markdown and PDFs

Lexomni MCP transforms your project's /_lexomni folder into a local knowledge base for AI agents. It indexes Markdown and PDF files using SQLite FTS5 for fast search, chunked reading, and auditable incremental memory.

Stop losing context between prompts. Give your agents long-term memory.

TypeScriptNode.jsMCPSQLite FTS5MarkdownPDF

One-Click Cursor Install

Works with CursorRuns locallyNo cloud required

mcp.json

{
  "mcpServers": {
    "lexomni": {
      "command": "npx",
      "args": [
        "-y",
        "lexomni-mcp"
      ],
      "env": {}
    }
  }
}

Cursor required

Star & Contribute

github.com/PagansDev/lexomni-mcp

Install & Use

npmjs.com/package/lexomni-mcp

WHY LEXOMNI EXISTS

Software projects accumulate critical knowledge: architecture decisions, guidelines, ADRs, internal patterns, and technical references that shape how a system evolves. This knowledge is often scattered across PDFs, Markdown files, wikis, and personal notes — hard to consult, easy to ignore, and rarely integrated into the real development workflow. Lexomni exists to turn knowledge into direction. It transforms the _lexomni folder into a local, searchable knowledge base that defines the project's technical context — for both humans and AI agents. With Lexomni, you can: • make guidelines and ADRs an active part of development • preserve architectural decisions over time • document business rules and internal standards • quickly consult technical reference materials and books • guide AI agents using sources you deliberately selected This includes everything from project documentation to iconic software engineering literature. You decide which ideas shape your system. You define the technical standard your AI should follow. You keep knowledge alive and versioned alongside the codebase. Because code changes. Architecture evolves. Knowledge must evolve with it. Lexomni doesn't just store information. It establishes technical direction.

Available tools

lexomni_listSources

Lists documents (MD and PDF) in the workspace

lexomni_buildIndex

Indexes documents in SQLite FTS5 for fast search

lexomni_searchDocs

Keyword search in indexed documents

lexomni_readDoc

Reads a specific chunk of a document

lexomni_writeNote

Creates or updates markdown notes in _lexomni/agent

Agent reads docs

Index builds

Search retrieval

Memory persists

Features

Structure /_lexomni: user/, agent/, books/, index/

Fast and accurate FTS5 search

Security: path traversal blocking, restricted access

Multilingual strategy with living glossary