Ogham MCP: Persistent Memory for AI Agents via Postgres

Ogham MCP offers a solution for AI agents that struggle with retaining information over time. It leverages a PostgreSQL database to store memories, employing a hybrid search approach that combines vector similarity and keyword matching for efficient retrieval. The system incorporates cognitive scoring mechanisms, inspired by ACT-R, to rank memories based on frequency of access, recency of use, and interconnectedness within the knowledge graph. This ensures that the most relevant and reliable information is prioritized. Profile isolation allows for context-specific memories, preventing knowledge leakage between different projects or tasks. The open-source nature of Ogham MCP, under the MIT license, encourages community contributions and customization. It supports various embedding providers, including OpenAI, Mistral, Voyage AI, and Ollama, offering flexibility in terms of privacy and performance. The use of a standard PostgreSQL database eliminates the need for specialized graph databases or LLMs in the write path, simplifying the architecture and reducing computational overhead. The hybrid retrieval approach, combining vector similarity and keyword matching, ensures that memories can be found both by meaning and by exact terms. The cognitive scoring system, incorporating factors such as access frequency, recency, and graph centrality, helps to surface the most relevant and reliable memories.

Transparency note: The analysis is based on the provided description of Ogham MCP and its features. Further evaluation is needed to assess its performance and scalability in real-world applications.

_Context: This intelligence report was compiled by the DailyAIWire Strategy Engine. Verified for Art. 50 Compliance._