DevMemory adds RAG skill for code agents
DevMemory shipped a RAG skill for coding agents that stores real code solutions across projects with live doc validation to fight prototype 'amnesia' in scaled agents — the feature was demoed announced on social. The approach pairs persistent example storage with runtime checks to keep coding agents accurate across multi-repo contexts.
Version 0.1.0 was published to PyPI on Feb 13, 2026, marking the package's first public release on the Python package index. pypi.org The runtime surfaces as a Model Context Protocol (MCP) server implemented in C#, with local deployment and explicit support for SQLite, PostgreSQL and Neo4j storage backends under an MIT license. mcpworld.com Package metadata states the system is built on Git‑capture tooling and integrates with a Redis‑based Agent Memory Server for semantic search and recall. pypi.org The project's GitHub describes commit- and prompt‑level provenance tracking so the memory can answer “why” a change was made by linking code lines to commits, prompts and context. github.com A local‑first VS Code extension and community forks highlight an offline deployment model that preserves source privacy and avoids sending repository contents to external services. marketplace.visualstudio.com Project pages and listings flag the offering as beta with a cloud mode “coming soon,” positioning the stack as a lightweight, team-oriented memory layer for agent pipelines during early production trials. github.com
