OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
RAG & Knowledge Systems

MCP Local Rag

MCP Local RAG is a TypeScript local RAG server that indexes supported documents on a machine and searches them by semantic similarity and keywords through an MCP client or CLI.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

MCP Local Rag — OpenViST project profile in RAG & Knowledge Systems.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you have supported local documents to search and want a semantic-plus-keyword retrieval workflow through an MCP client or terminal.

Who it is for

  • Developers who want to search private technical documentation or code-related reference material locally.
  • People using an MCP client who want document search as part of their client workflow.
  • Terminal users who want to ingest and query a local document index through a CLI.

Documented uses

  • Find documentation passages relevant to a question while also matching exact technical terms.
  • Search local PDF, DOCX, Markdown and text collections without sending documents to an embedding API.
  • Inspect surrounding chunks and use available source locations to check a result against its original document.
  • Refresh an index after documents are added, edited or removed.
Before you choose

Requirements & limits

Requires Node.js 22 or later. First use requires internet access to download the npm package and embedding model; subsequent text ingestion and search work offline after the model is cached. The repository identifies the project as MIT-licensed.

  • Combines semantic similarity with keyword matching rather than relying only on one search method.
  • Indexes PDF, DOCX, TXT and Markdown files; HTML must be fetched by the client before it is passed to the server.
  • Requires an explicit document root for MCP use; BASE_DIR is also the file-operation boundary.
  • Search results can include passages and, where available, headings, line numbers or page numbers.
  • Sync skips unchanged files and removes entries for deleted files; repeat sync after collection changes.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Set up an MCP document index

Configure an MCP client to index a local directory and search its supported documents.

Before you start

  • Node.js 22 or later
  • Internet access on first use to download the npm package and embedding model
  • A directory of documents and its absolute path
  • An MCP client with a configuration format supported by the project

Steps

  1. Set BASE_DIR to the absolute path of the document directory in the MCP server configuration.

  2. Register the command below as the local MCP server using the configuration format for your client.

    Documented command
    npx -y mcp-local-rag

  3. Restart the client and request a sync of all documents in the configured root.

  4. After sync completes, ask the client a question about the indexed documents.

Expected outcome

The MCP client can search the indexed local documents using semantic matching and keyword search.

Guide limitations

  • The first sync downloads the default embedding model, about 90 MB, and may take 1–2 minutes before ingestion starts.
  • MCP file paths must be absolute and inside a configured document root.
  • Supported file ingestion covers PDF, DOCX, TXT and Markdown; HTML must already have been fetched by the client.
Documentation 1 ↗
Source-based practical guide

Ingest and query documents with the CLI

Use the terminal to ingest a document directory and search its index.

Before you start

  • Node.js 22 or later
  • Internet access on first use to download the npm package and embedding model
  • A directory containing supported documents

Steps

  1. Open a terminal in the document directory.

  2. Run the command below to ingest the selected directory.

    Documented command
    npx mcp-local-rag ingest ./docs/

  3. Run the command below to search it.

    Documented command
    npx mcp-local-rag query "authentication API"

Expected outcome

The CLI returns search results from the ingested documents, matching semantic relevance and keywords.

Guide limitations

  • The CLI uses the current directory as its document root by default; run ingestion and query from the same directory or set BASE_DIR and DB_PATH explicitly.
  • File ingestion supports PDF, DOCX, TXT and Markdown, not Excel, PowerPoint, standalone images or source-code file extensions.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
13 October 2026
Original website title
shinpr/mcp-local-rag · Official homepage title · retained in its source language
How we review sourcesSuggest a correction