Show HN: Self-improving skills for any coding agent

Category: infrastructure

Tags: mcp-server, agent-memory, ai-assistant

Score: 7.0/10 (Innovation: 7, Technical: 7, Documentation: 7, Utility: 7)

Dreamer is a self-hosted MCP server that collects memories from coding agents and periodically regenerates a shared context bundle, enabling team-wide persistent knowledge for AI assistants. Its plugin architecture via Python protocols and YAML configuration makes it extensible, and the focus on agent memory pooling is a novel approach to improving coding agent consistency.

Target audience: backend devs, devops, ai engineers

Repository: https://github.com/luml-ai/dreamer · Python · MIT · 27 stars

View on Hacker News