Cog-Bot

Available

Cognitive memory for AI agents

AI agents forget everything between conversations. Cog-Bot fixes that — it gives agents persistent, evolving memory so they build on what they've learned instead of starting from scratch every time.

Cog-Bot solves the fundamental problem with AI agents: they forget everything between conversations. Built on Spring Boot 3.4 WebFlux, it provides agents with persistent memory organized into cognitive spaces — semantic clusters that develop their own health scores and knowledge hierarchies.

Memory flows through a multi-stage pipeline: distill, embed, save, link, evolve. A background heartbeat manages centroid refresh, tier migration (HOT to WARM to COLD), health checks, and weekly curation sweeps. The system discovers bridges between spaces, detects knowledge gaps, and synthesizes cross-domain insights automatically.

Designed for production: reactive throughout, MCP-compatible, with a trust ledger for agent autonomy management.

Key Features

  • Semantic search via pgvector with multi-tier memory lifecycle
  • Cognitive spaces with centroid embeddings and organic health scoring
  • Automatic knowledge curation: dreaming cycles, bridge discovery, synthesis
  • Agent daemon with goal-cycle execution (GOAL, PLAN, EXECUTE, REFLECT, LEARN)
  • MCP-compatible tool interface for any AI model
  • Human-in-the-loop workbench for task delegation and oversight
  • Trust ledger with tiered autonomy based on agent track record

Built With

Java 21Spring Boot 3.4Spring AIWebFluxpgvectorPostgreSQLMCP

Interested in this solution?

I can help you deploy, customize, or build something similar for your specific needs.