Skip to main content
MCP server CX Codex

Memnest — Claude Code MCP server: memory and setup

The Memnest MCP server for Claude Code stores decisions and retrieves them from earlier conversations. An agent can query the shared local memory to recover the reason behind a choice.

MIT ★ 6 updated 2026-11-14

Open on GitHub

What Memnest does

Memnest keeps conversation text and decisions so connected coding agents can retrieve them later. Search combines matching terms with semantic similarity, and each result can be opened through its ID. A session watcher can also save visible turns from Claude Code, Codex, and pi.

Project folders have separate memory scopes, while playbook is shared across them. The server offers tools to search, read, add, edit, and delete entries, plus task cards that carry work into a later session. Records live in SQLite on the machine; saving and searching do not call an API.

Supported systems are macOS on Apple Silicon, Linux with systemd, and 64-bit Windows 10/11. Python 3 is required. The first save or search fetches a local model of roughly 1.1 GB; on Windows, Python connects the clients.

What it helps with

  • recover a decision from an earlier conversation
  • open the original text behind a search result
  • share saved decisions across connected agents
  • resume a task card in a later session

Example prompts for the agent

  • Find why we chose the current library and open the original discussion.
  • Save this decision: migrations run only from the release branch.
  • Create a task card for the migration so work can continue in another session.

How to install Memnest

Install

The MCP registry has no launch command for it: see the repository README

Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more

Questions about Memnest

How do I install Memnest in Claude Code?

Run curl -fsSL https://raw.githubusercontent.com/Blue-B/memnest/main/core/scripts/install.sh | bash in a terminal. The installer starts the server and conversation watcher, then connects Claude Code, Codex, and Cursor configurations it finds.

What does Memnest do?

It stores conversation text and decisions locally, then retrieves entries through keyword and semantic search. A connected agent can open the original record and refer to its ID.

What does Memnest need to run?

It requires macOS on Apple Silicon, Linux with systemd, or 64-bit Windows 10/11, plus Python 3. The first save or search downloads a local model of about 1.1 GB; on Windows, Python is used to connect clients.

Similar MCP servers

All MCP servers for Agent setup: memory, context, orchestration →

Project sponsors

Companies that keep this analytics open