Selvedge

What is this MCP

Selvedge is a Model Context Protocol (MCP) server that provides persistent decision memory for AI agents. It allows AI agents to explicitly save their decision-making rationales, rejected approaches, entity history, and conditions for reconsideration in a local SQLite database. The system is designed to maintain context across sessions without making LLM calls for core storage and retrieval operations.

How to use this MCP

To use Selvedge, you need to run the MCP server locally via the CLI using uvx --from selvedge selvedge-server. The server communicates over stdio protocol, making it compatible with various AI agents that support MCP. Once running, compatible coding agents and developers can connect to it to save and retrieve decision memory across different sessions.

What this MCP can be used for

This MCP is primarily used for enhancing AI agent capabilities by providing persistent memory of past decisions. It enables agents to remember why certain approaches were chosen or rejected, maintain entity histories, and track conditions that might warrant reconsideration of previous decisions. This is particularly valuable for long-running AI projects, coding assistants, and any scenario where maintaining decision context across sessions improves agent performance and consistency.