CogniCore

Overview

CogniCore is a community observability tool for AI agents that provides persistent memory, reflection, and safety. It uses BM25 and multi-hop graph retrieval without requiring a vector database or embedding server. It includes an MCP server with 30 tools.

Best for

Best for
Developers building AI agents who want lightweight persistent memory and safety without vector infrastructure.

Use cases

• Add persistent memory to AI agent workflows
• Retrieve relevant context using BM25 and graph traversal
• Integrate reflection and safety checks via MCP tools

Notes

CogniCore is a community observability tool for AI agents that provides persistent memory, reflection, and safety. It uses BM25 and multi-hop graph retrieval without requiring a vector database or embedding server. It includes an MCP server with 30 tools.

73 stars on GitHub. Last updated 2026-10-02. Licensed MIT.

Use cases

• Add persistent memory to AI agent workflows
• Retrieve relevant context using BM25 and graph traversal
• Integrate reflection and safety checks via MCP tools

Pros

• Zero dependencies simplify deployment
• No vector DB or embedding server needed
• MCP server offers 30 tools for agent integration

Cons

• Low star count suggests limited community adoption
• Python only may restrict language portability
• Observability scope may overlap with existing tracing tools

Indexed from awesome-llmops and enriched against its public facts.