Log inSubmit server
Rvf MCP Server logo

Rvf MCP Server

by @ruvector

Local MCP access to portable RVF vector stores for creating, populating, and querying HNSW indexes.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install Rvf MCP Server

Pick your client, then add it in one click or paste the config.

STDIONode.js
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "ruvector-rvf-mcp-server": {
      "command": "npx",
      "args": [
        "-y",
        "ruvector"
      ],
      "env": {
        "RUVECTOR_MCP_PROFILE": "YOUR_VALUE_HERE"
      }
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Values are written into the snippet above so you can paste it as is. They never leave your browser.

Install not yet sandbox-tested— we haven't run this command through our automated sandbox yet. Not a red flag; we're working through the catalog.

Overview

@ruvector/rvf-mcp-server exposes local vector search over portable RVF files to MCP-compatible clients. It lets agents create, populate, and query HNSW-indexed stores through stdio or SSE transports. The server is included in the ruvector npm package and can be started with npx. Reach for it when an agent needs project-local vector retrieval or persistent memory without relying on a hosted database.

Use cases

  • Create project-local RVF vector stores
  • Populate persistent agent memory
  • Query HNSW indexes for relevant context
  • Keep vector retrieval on a local machine

Key features

  • Portable RVF file storage
  • HNSW-indexed vector search
  • MCP tools over stdio or SSE
  • Read-only operating profile
  • npm and npx-based startup

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 8h ago via OSV.dev · ruvector (npm)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars4.7kStargazers on the source repository.
npm downloads1.7kPackage downloads in the last 30 days.
Last commitTodayMost recent push to the default branch.

How Rvf MCP Server works

What @ruvector/rvf-mcp-server does

@ruvector/rvf-mcp-server gives MCP clients access to local vector stores saved as portable RVF files. Agents can create stores, add records, and run vector queries against HNSW indexes. The local storage model suits project memory, semantic retrieval, and other workflows where vector data should remain available in the project environment.

The server is part of the ruvector npm package. It is intended for MCP-compatible agent applications, including Claude Code and Cursor. Its tools can be exposed over standard input/output or through SSE, depending on how the server is launched and connected.

How it works

An MCP client starts the server process and communicates with it using the selected transport. For a local stdio setup, the client launches npx with ruvector mcp start. The project directory acts as the working directory, allowing the server to use the associated local RuVector data.

The package includes an MCP tool inspection command, npx ruvector mcp tools, which can be used before connecting a client. A read-only operating profile is available through the RUVECTOR_MCP_PROFILE environment variable. Write access is intended to be enabled deliberately through an explicit tool policy rather than assumed by the client.

Setup and configuration

Install Node.js and npm, then use the ruvector package through npx or install it in the project. A basic read-only launch is:

bash
RUVECTOR_MCP_PROFILE=readonly npx ruvector mcp start

Configure the MCP client to run npx with the arguments ruvector mcp start, set the working directory to the project, and pass the profile environment variable when needed. Claude Code can register a project-scoped stdio connection with:

Terminal
claude mcp add --scope project --env RUVECTOR_MCP_PROFILE=readonly --transport stdio ruvector -- npx -y ruvector mcp start

After registration, use the client's MCP status or connection view to verify that the server is available. Keeping the npm lockfile in the project helps preserve the selected dependency version.

Tools and capabilities

  • Inspect the available MCP tools before connecting an agent.
  • Create local RVF vector stores.
  • Fill stores with records for later retrieval.
  • Query HNSW-indexed vector data.
  • Expose local vector search through stdio or SSE.
  • Apply a read-only profile for safer agent access.

Limitations and notes

The server depends on a local Node.js and npm environment. Semantic text workflows also require an embedding model; the first semantic command in the broader ruvector package downloads a local model. Retrieved records should be treated as project evidence and checked against current source files when used as agent memory.

Write operations should be governed explicitly through the MCP client's tool policy. A read-only profile is the documented starting point for connecting an agent, while populating memory can be performed separately through the npm-based RuVector workflow or an intentionally enabled write policy.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Rvf MCP Server

@ruvector/rvf-mcp-server is an MCP server that connects AI clients to local portable RVF vector files. Its main tools create stores, fill them with records, and query HNSW indexes. It exposes these operations over stdio or SSE for clients such as Claude Code and Cursor.

Run it through the `ruvector` npm package with `npx`; a separate server installation is not required. Start it with `RUVECTOR_MCP_PROFILE=readonly npx ruvector mcp start`, or install `ruvector` in the project and configure the MCP client to launch the same command.

Yes, it works with Cursor as an MCP client. Configure Cursor to launch `npx` with `ruvector mcp start`, provide the appropriate environment settings, and use the project directory as the working directory.

Set `RUVECTOR_MCP_PROFILE=readonly` before starting the server. This profile is intended as the initial connection mode; write access should be enabled separately through an explicit MCP tool policy.

Yes, the server supports SSE as well as stdio. Select the transport supported by the MCP client and configure the connection accordingly; the documented local setup uses stdio with `npx ruvector mcp start`.

Related MCP servers

All in Developer Tools All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewRvf MCP Server AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/ruvector-rvf-mcp-server?style=directory)](https://allmcps.com/mcp/ruvector-rvf-mcp-server)
HTML Embed
<a href="https://allmcps.com/mcp/ruvector-rvf-mcp-server"><img src="https://allmcps.com/api/badge/ruvector-rvf-mcp-server?style=directory" alt="Rvf MCP Server on AllMCPs" /></a>