kratos
🏛️ Memory System for AI Coding Tools - Never explain your codebase again. MCP server with perfect project isolation, 95.8% context accuracy, and the Four Pillars Framework.
claude mcp add --transport stdio ceorkm-kratos-mcp npx --yes kratos-mcp@latest
How to use
Kratos MCP provides an ultra-lean memory system that helps AI tools remember project context across sessions. It offers a compact set of 12 tools focused on memory management, security, project management, and system status, all designed to minimize context overhead while preserving important knowledge about your codebase, projects, and decisions. To use Kratos, you typically run it as an MCP server (via npx kratos-mcp or a global installation) and then configure your preferred AI tool (Claude, Cursor, Windsurf, Continue, or any MCP-compatible client) to connect to the server. Once configured, you can save memories, search them semantically, retrieve specific memories, and perform lightweight project management tasks—all while keeping data local in per-project SQLite databases. The Quick Start section in the README shows common setup patterns for Claude Desktop, Claude Code, and Cursor to integrate Kratos into your workflow. Kratos emphasizes zero configuration and auto-detection of relevant projects, with secure, private storage on your machine.
How to install
Prerequisites:
- Node.js and npm (or npx available via npm).
- Basic familiarity with MCP (Model Context Protocol) tooling and your AI integration.
Installation steps:
- Install Kratos MCP globally (optional):
npm install -g kratos-mcp
- Run Kratos MCP directly with npx (no installation required):
npx kratos-mcp
- Or install as a project dependency in your app/repo:
npm install kratos-mcp
- If you need to customize the server quickly, use the recommended npx invocation shown in the README configurations:
npx --yes kratos-mcp@latest
- Ensure your AI tool is configured to point at the Kratos MCP server using the mcp_config example provided in the README (or through your tool’s MCP settings).
Additional notes
Tips and notes:
- Kratos stores data locally per project in a SQLite database, ensuring isolation between projects.
- The default storage location is ~/.kratos/ with a projects/ subfolder containing individual memories.db files.
- Kratos auto-detects your project based on git, package.json, or directory structure, reducing setup time.
- If you need to relocate data, Kratos supports changing the storage path (as noted in the README v1.6.1 update).
- The MCP configuration typically uses npx kratos-mcp@latest as the command, which keeps you aligned with the latest features and security fixes.
- When integrating with tools like Claude Desktop, Claude Code, or Cursor, you can copy-paste the provided mcpServers snippet into your tool’s config to enable seamless memory-backed interactions.
- If you encounter issues, verify that your environment has the required Node/NPX tooling and that the MCP server is reachable from your AI tool (network and firewall settings permitting).
Related MCP Servers
sandboxed.sh
Self-hosted orchestrator for AI autonomous agents. Run Claude Code & Open Code in isolated linux workspaces. Manage your skills, configs and encrypted secrets with a git repo.
task-orchestrator
A light touch MCP task orchestration server for AI agents. Persistent work tracking and context storage across sessions and agents. Defines planning floors through composable notes with optional gating transitions. Coordinates multi-agent execution without prescribing how agents do their work.
claude-code-open
Open source AI coding platform with Web IDE, multi-agent system, 37+ tools, MCP protocol. MIT licensed.
lc2mcp
Convert LangChain tools to FastMCP tools
Pare
Dev tools, optimized for agents. Structured, token-efficient MCP servers for git, test runners, npm, Docker, and more.
local -gateway
Aggregate multiple MCP servers into a single endpoint with web UI, OAuth 2.1, and profile-based tool management