Continual Learning System for AI Agents
Give your AI agents memory and skills that compound over time
Phronais is a Model Context Protocol (MCP) server that enables AI agents to learn from every interaction, extract reusable patterns, and apply accumulated knowledge across sessions.
Store observations with context, outcomes, and metadata. Semantic search retrieves relevant past experiences.
Extract reusable patterns from memories and codify them as skills. User-approved skills compound your agent's capabilities.
Automatic pattern detection suggests new skills based on repeated successful approaches. Your agent gets smarter over time.
Self-extending capability system. Create custom MCP tools, HTTP proxies, transforms, and skill-backed functions.
OAuth 2.1 authentication with Google. Rate limiting, storage quotas, and user-scoped data isolation.
Run locally via MCP stdio or deploy to Cloudflare Workers. Works with Claude Desktop, Claude.ai, and Claude Code.
Claude stores observations from your work with context about what happened and what worked.
The system detects patterns in your interactions and suggests reusable skills.
Approved skills are automatically suggested in relevant contexts, making Claude more capable over time.
Connect via MCP stdio in your config
OAuth authentication for cloud access
Full MCP integration for coding sessions
Get started in under 5 minutes with our quickstart guide.