Codemem by Cogniplex is a free MCP server tool designed to provide durable, local-first memory for AI coding agents and developers. Built primarily with Rust and TypeScript, it enhances AI coding assistants such as OpenCode, Claude Code, Cursor, and Windsurf by offering advanced memory capabilities. Codemem ensures project context and architectural decisions are retained, moving beyond transient chat transcripts. It integrates into project environments to manage lifecycle hooks and configure the MCP server, supporting various distribution methods for easy access. This tool aims to improve the effectiveness of AI in real engineering workflows.

What does Codemem connect to?

Codemem is an MCP server tool that connects to and enhances various AI coding assistants, including OpenCode, Claude Code, Cursor, and Windsurf. It is also confirmed compatible with AI clients like Codex CLI. This tool establishes a crucial link for AI agents to access and manage project-specific memory. It supports stdio as a JSON-RPC tool bridge and streamable HTTP for local transport testing, ensuring flexible integration with different client architectures and communication methods.

Core features and tools

Codemem exposes 32 distinct MCP tools to AI assistants, significantly expanding their capabilities. These include specialized functions such as temporal queries, diff review, and blast radius analysis. Additionally, it offers 5 dedicated temporal MCP tools, focusing on time-sensitive data and history. These tools collectively enable AI agents to perform more sophisticated analyses and retain context over extended periods, moving beyond simple conversational memory. Its architecture supports durable project memory for real engineering work.

Who is it for?

Codemem is built for AI coding agents and individual developers or engineers seeking to enhance their AI assistant's memory and contextual understanding. It provides solutions for retaining architecture decisions, recovering debugging context, and transferring knowledge between humans and agents. This tool also helps remember repository-specific issues without needing external wikis. Skip this if your workflow does not involve AI coding agents or require durable project memory for engineering tasks.

How to get started

For local use, codemem's setup involves an initialization process within a project. For more advanced or remote MCP server authentication, Access is granted through a standard OAuth authorization flow, potentially via a proxy like CodeMie, supporting methods like CodeMie SSO, JWT Bearer Authorization, and API Key authentication. OIDC configuration can be set via environment variables.

Use cases

• Recovering debugging context after a dead-end session.
• Retaining architecture decisions linked to their creation.
• Facilitating context transfer between humans and coding agents.
• Remembering repository-specific issues without needing a separate wiki.

Codemem by Cogniplex provides essential memory capabilities for AI coding agents and developers. Its ability to expose numerous MCP tools, including temporal queries and diff review, coupled with support for various AI clients, makes it a valuable addition to AI-powered development workflows. The tool aims to create durable project memory, moving past ephemeral chat transcripts. Start by experimenting with its context recovery features after a debugging session.