Skip to content
MCP server

Tzivia-AI

Persistent knowledge-graph memory for AI agents, hosted or on a local Windows install

Install
Platforms
Windows, API, self-hosted
Runs
Locally or Remotely
Version
5.1.1
Listing updated
today

Similar MCP servers

About Tzivia-AI

Tzivia-AI is an MCP server for knowledge and memory. It can run locally on your machine or as a hosted remote server.

Rather than storing memory as a flat vector pile, Tzivia builds a typed, decaying knowledge graph. Its ARIA routing algorithm walks the graph without scanning every item, nodes are scored on frequency, accuracy, connectivity and inference weight (Vec4), and ten typed edge types let a logic engine reason over relationships. A synthesis layer that does not decay keeps discovered patterns instead of recomputing them each session.

The project reports its own benchmarks: routing 3,136 times faster than linear search at 100,000 items, 88 ms to retrieve 25 knowledge chains from a 1,000,000-token corpus, and a 99.6% search-space reduction. The README says methodology is available on request, and public LongMemEval and LoCoMo results are described as coming.

It can be used as a hosted connector at mcp.tzivia.dev or as a local install packaged as an mcpb bundle. The README describes local storage in a single SQLite file and says it is single-tenant, with a hosted multi-tenant PostgreSQL backend on the roadmap. The README also describes a research use of the same engine, a DNA decay finder validated against the public TCGA dataset, and states it is not a clinical tool.

Install

Run in a terminal

claude mcp add --transport http tzivia https://mcp.tzivia.dev/mcp

Add to Settings → Connectors → Add custom connector

https://mcp.tzivia.dev/mcp
Install in CursorOpens Cursor, which asks you to confirm. Or add it by hand:

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "tzivia": {
      "url": "https://mcp.tzivia.dev/mcp"
    }
  }
}
Install in VS CodeOpens VS Code, which asks you to confirm. Or add it by hand:

Add to .vscode/mcp.json

{
  "servers": {
    "tzivia": {
      "type": "http",
      "url": "https://mcp.tzivia.dev/mcp"
    }
  }
}

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

{
  "mcpServers": {
    "tzivia": {
      "serverUrl": "https://mcp.tzivia.dev/mcp"
    }
  }
}

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

{
  "mcpServers": {
    "tzivia": {
      "type": "http",
      "url": "https://mcp.tzivia.dev/mcp"
    }
  }
}

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

What it does

Typed knowledge graph
Memory modelled with ten kinds of relationship edges.
ARIA routing
Sub-linear traversal that avoids scanning the full store.
Vec4 scoring
Nodes scored on frequency, accuracy, connectivity and inference.
Synthesis layer
Durable insights that do not decay.
Evidence trail
Recall returns the supporting evidence.
Local SQLite storage
The local build keeps the graph in one file.

Good for

  • Give an agent memory that persists between sessions
  • Cut tokens spent re-reading large corpora
  • Recall a precise fact or a distilled principle on demand
  • Keep a private memory graph on your own machine

Pricing

Pricing unknown

Prices not published

Prices change. Check tzivia.dev before you buy.

Tzivia-AI FAQ

Where does my memory live?

In the hosted connector or, for the local install, in a single SQLite file on your machine.

Are the benchmark figures independent?

They are the project's own results; methodology is available on request.

What is the DNA work?

An early research application, not a validated clinical tool.

Which clients can use it?

Claude, Cursor and any MCP-compatible host.

How do I add Tzivia-AI to Claude Code?

Run this command in your terminal: claude mcp add --transport http tzivia https://mcp.tzivia.dev/mcp

What are the best alternatives to Tzivia-AI?

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

Where this comes from

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

How AIPoweredBot checks servers