MemoryMesh

The SQLite of AI Memory. Persistent, intelligent memory for any LLM in three lines of code.

MIT License 100% Free Zero Dependencies Fully Local Any LLM
pip install memorymesh
Private 100% local storage
Persistent Memory across sessions
Zero Dependencies to install
3 Lines To get started

Three Lines. Persistent Memory.

from memorymesh import MemoryMesh

memory = MemoryMesh()
memory.remember("User prefers Python and dark mode")
results = memory.recall("What does the user prefer?")

No servers. No cloud. No API keys. Everything stays on your machine in SQLite.

Works With Everything

Claude Code Claude Desktop GPT-4 Gemini CLI Cursor Windsurf Ollama Codex CLI Llama Mistral GitHub Copilot Any MCP Client
npx -y @smithery/cli install @sparkvibe-io/memorymesh --client claude

Or connect at smithery.ai/servers/sparkvibe-io/memorymesh. Supports 20+ MCP clients.

{ "mcpServers": { "memorymesh": { "command": "memorymesh-mcp" } } }

Works with Claude Code, Gemini CLI, Codex CLI, Cursor, Windsurf, and any MCP-compatible tool.

from memorymesh import MemoryMesh
memory = MemoryMesh(embedding="ollama")
memory.remember("User prefers dark mode")

Why MemoryMesh?

AI tools have no memory. Existing solutions are heavy. MemoryMesh is different.

Sessions Start Blank

You told Claude your coding style yesterday. Today it asks again.

Decisions Are Lost

You decided JWT for auth. Next session, the AI suggests cookies.

Mistakes Repeat

Fixed a bug Monday. Friday, the AI introduces the same bug.

Tools Are Siloed

What you teach Claude stays in Claude. Gemini and Cursor know nothing.

SolutionApproachTrade-off
Mem0SaaSCloud required, data leaves machine, ongoing costs
LettaAgent frameworkHeavy lock-in, complex setup
ZepMemory serverRequires PostgreSQL + Docker
MemoryMeshEmbeddable libraryZero deps. SQLite. Free forever.

Features

Simple API

remember(), recall(), forget(). No boilerplate.

Semantic Search

Find by meaning, not keywords. Ollama or local embeddings.

MCP Server

Built-in for Claude, Cursor, Windsurf, Gemini.

Memory Categories

Auto-categorize and route to the right scope.

Encrypted Storage

Encrypt at rest. Zero external deps.

Session Context

User profile, guardrails, project state in one call.

Privacy Guard

Auto-detect and redact secrets before storing.

Cross-Tool Sync

Sync memories across Claude, Codex, Gemini, Cursor.

What You Actually Get

MemoryMesh isn't about cramming more into the context window. It's about making every session smarter from the start.

Without MemoryMeshWith MemoryMesh
Every session starts blankPreferences, decisions, and patterns load automatically
Repeat yourself across toolsOne memory store shared across Claude, Gemini, Codex, Cursor
AI re-discovers settled decisionsPast decisions recalled by semantic search
Same bugs reintroducedPast mistakes remembered and avoided
Context locked in one toolPortable memory that follows you

Give Your AI a Memory

pip install memorymesh