Skip to content
MCP server

VelesDB Memory

Local agent memory server with remember, recall, relate, forget and why over one engine

Runs locallyStatus and history
Install
Platforms
macOS, Windows, Linux
Runs
Locally
Version
0.17.1
Listing updated
today

Similar MCP servers

About VelesDB Memory

VelesDB Memory is an MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.

Memory lives in a single local engine that fuses vector search, a graph and columnar storage, so one store can answer similarity questions and relationship questions together. The README says no LLM and no API key sit in the memory path, and data stays on the machine. The only outbound call is an opt-out version check that sends version, edition, OS, architecture and a hashed install id.

The registry lists five operations: remember, recall, relate, forget and why. The last one is the notable piece: the README describes a memory that shows its evidence rather than returning bare matches. Prebuilt .mcpb bundles are published for Apple Silicon and Intel macOS, Windows x64 and Linux on x86_64 and ARM, and the server runs over stdio.

A guided installer detects installed clients (Claude Code, Codex, Cursor, Cline, Zed, OpenCode, Claude Desktop, Windsurf and others), picks single-client stdio or a shared local daemon, and can set up optional hooks that load and save working context and require a recall before an edit. Embeddings can come from a local Ollama model; the embedding dimension is fixed per store.

Install

Run in a terminal

claude mcp add velesdb-memory -- npx -y https://github.com/cyberlife-coder/VelesDB/releases/download/velesdb-memory-v0.17.1/velesdb-memory-0.17.1-aarch64-apple-darwin.mcpb

Add to claude_desktop_config.json

{
  "mcpServers": {
    "velesdb-memory": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cyberlife-coder/VelesDB/releases/download/velesdb-memory-v0.17.1/velesdb-memory-0.17.1-aarch64-apple-darwin.mcpb"
      ]
    }
  }
}
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "velesdb-memory": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cyberlife-coder/VelesDB/releases/download/velesdb-memory-v0.17.1/velesdb-memory-0.17.1-aarch64-apple-darwin.mcpb"
      ]
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "velesdb-memory": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cyberlife-coder/VelesDB/releases/download/velesdb-memory-v0.17.1/velesdb-memory-0.17.1-aarch64-apple-darwin.mcpb"
      ]
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "velesdb-memory": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cyberlife-coder/VelesDB/releases/download/velesdb-memory-v0.17.1/velesdb-memory-0.17.1-aarch64-apple-darwin.mcpb"
      ]
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "velesdb-memory": {
      "command": "npx",
      "args": [
        "-y",
        "https://github.com/cyberlife-coder/VelesDB/releases/download/velesdb-memory-v0.17.1/velesdb-memory-0.17.1-aarch64-apple-darwin.mcpb"
      ]
    }
  }
}

Using several servers? Build one config file with the MCP config generator.

Tools this server provides (5)

Taken from the project documentation; the running server may differ.

remember
Stores an item in the agent memory.
recall
Retrieves stored items relevant to a query.
relate
Links stored items to each other in the graph.
forget
Removes stored items from memory.
why
Shows the evidence behind a recalled item.

What it does

Fused vector, graph and columns
One binary of about 14 MB combines three storage models under a single query language.
Evidence-backed recall
The why operation explains where a remembered item came from.
Local-first
No LLM or API key is used in the memory path and data is not sent out.
Guided installer
A wizard detects supported clients, checks checksums and rewires configs with backups.
Optional agent hooks
Six Claude Code hooks save working context and can require recall before edits.
Shared local daemon
One memory can serve several clients over local HTTPS.

Good for

  • Give a coding agent persistent memory across sessions
  • Share one memory between Claude Code and Cursor on one machine
  • Keep related facts linked so an agent can recall them by relationship
  • Run agent memory offline with a local embedding model

VelesDB Memory FAQ

Does it send my data anywhere?

The README says data never leaves the machine; the only outbound call is an opt-out version check with no stored data.

Which platforms have prebuilt bundles?

macOS on Apple Silicon and Intel, Windows x64, and Linux on x86_64 and ARM.

Can I install it without Rust?

Yes, via a prebuilt .mcpb bundle or the npm package @wiscale/velesdb-memory-node.

What is the embedding dimension tied to?

The embedding model; it is fixed when the store is created, and changing it needs a re-embed.

How do I add VelesDB Memory to Claude Code?

Run this command in your terminal: claude mcp add velesdb-memory -- npx -y https://github.com/cyberlife-coder/VelesDB/releases/download/velesdb-memory-v0.17.1/velesdb-memory-0.17.1-aarch64-apple-darwin.mcpb

What are the best alternatives to VelesDB Memory?

The closest alternatives to VelesDB Memory are Codebase Memory, XERJ, Claude Code Ultimate Guide, Nautilus Compass Memory and NeuroStack.

Where this comes from

  • Description: written from the project's README.
  • Install steps: from the official MCP Registry entry.

How AIPoweredBot checks servers