AgentsMedium impactFor DevGitHub MCP Servers · May 18, 2026

Self-hosted MCP server for AI agent memory - 14 core tools, profile-based tiering (86+ available). 87.8% LoCoMo. Works with Claude, Cursor, Windsurf.

Dakera-AI/dakera-mcp

Dakera-MCP is a self-hosted server implementing the Model Context Protocol for AI agent memory, supporting over 86 tools and compatibility with models like Claude, Cursor, and Windsurf.
Signal strength3.8/5·1 stars

Dakera-MCP is a self-hosted server implementing the Model Context Protocol for AI agent memory, supporting over 86 tools and compatibility with models like Claude, Cursor, and Windsurf.

TL;DR

Dakera-MCP is a self-hosted server implementing the Model Context Protocol for AI agent memory, supporting over 86 tools and compatibility with models like Claude, Cursor, and Windsurf.

What happened

A GitHub repository was created for Dakera-MCP, a Rust-based MCP server designed to manage AI agent memory with profile-based tiering and multiple tools, facilitating improved context handling for AI agents.

Why it matters

Effective memory management is crucial for scalable, context-aware AI agents; a self-hosted MCP server allows customization and control over agent memory, enhancing multi-agent systems and integrations with diverse LLMs.

Generating deep dive...

AI-powered analysis takes a few seconds