Advertise
MCP.so
Sign In

PMB (Personal Memory Brain) | Local first memory layer MCP Serve…

@oleksiijko

Repository

Basic information

Category

Developer Tools

Transports

stdio

Author

oleksiijko

Submitted by

Oleksii Bondar

About PMB (Personal Memory Brain) | Local first memory layer MCP Server

Local-first persistent memory for AI coding agents over MCP. Gives Claude Code, Cursor and Codex shared, durable project memory across sessions. No cloud, no API keys.

#memory#ai-agents#mcp#local-first#rag#knowledge-graph#claude-code#cursor#codex#python

PMB (Personal Memory Brain) | Local first memory layer MCP Server config

Add this server to your MCP-compatible client using the configuration below.

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

PMB (Personal Memory Brain) | Local first memory layer MCP Server tools

Tool list not available yet

This listing does not currently have a tool list to display. This does not mean the server has no tools; check its documentation for supported capabilities.

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is PMB?

PMB (Personal Memory Brain) is a local-first persistent memory layer for AI coding agents over the Model Context Protocol. It provides Claude Code, Cursor, Codex, and other MCP-aware agents with a shared, durable memory of project decisions, lessons, facts, and goals that persists across sessions, stored entirely in a single SQLite file on your machine.

How to use PMB?

Install the package via pip (pip install pmb-ai), then connect it to your agent with a command like pmb connect claude-code. PMB works out of the box with Claude Code, Cursor, Codex, Windsurf, Zed, VS Code, and other MCP-compatible tools.

Key features of PMB

  • 100% local — no cloud, no API keys, no telemetry
  • Hybrid retrieval using BM25, vector search, and entity graph
  • Achieves 94.5% evidence-recall@10 on the LoCoMo benchmark
  • Persistent memory across all coding sessions
  • Compatible with multiple MCP-aware agents

Use cases of PMB

  • Retaining project decisions and rationale between coding sessions
  • Storing lessons learned and domain facts for consistent agent behavior
  • Enabling multiple AI agents to share and update the same project memory
  • Building a durable, searchable knowledge base for long-running development workflows

Frequently asked questions

1

Where is my data stored?

All memory is kept in a single SQLite file on your local machine. No data is sent to the cloud or any third party.

2

What agents does PMB work with?

PMB is designed for any MCP-aware agent, including Claude Code, Cursor, Codex, Windsurf, Zed, and VS Code.

3

Do I need an API key or internet connection?

No. PMB requires no API keys, no cloud services, and no internet connectivity. It is entirely local.

4

How do I install PMB?

Install the Python package with `pip install pmb-ai`, then run `pmb connect <agent-name>` (e.g., `pmb connect claude-code`) to link it to your agent.

5

What license is PMB released under?

PMB is released under the Apache-2.0 license.

Comments

More Developer Tools MCP servers

how.is — domain & website intelligence

how.is

Remote MCP server for domain and website checks: DNS, WHOIS/RDAP, domain age, TLS, DNSSEC, email auth (SPF/DKIM/DMARC), blocklists, a trust verdict ("should my agent click, sign in to, or pay this site?"), availability +

Leadpages MCP

Leadpages

Leadpages MCP is a remote MCP server available at https://mcp.leadpages.com.

Modelglass

Modelglass

Modelglass turns "which model should I use?" into a lookup instead of a guess. Seven tools give your agent live pricing across image, language, video, and audio models. Sourced and versioned, not scraped once and forgott

1

ZeroBeacon — 1000-Tool Collision-Proof Commerce Router

DavidFox998

ZeroBeacon.ai — Collision-proof commerce router for AI agents. 1000 MCP tools: commerce, trust, billing, sovereign ID + everyday tools. LIVE.

FARPY

mangomunchr

Verified distributed compute for humans, applications, and agents. Submit Blender, Octane, and video workloads, route them across qualified GPUs, track asynchronous execution, download finished artifacts, and verify FARP

1
Advertise