Mnemos MCP
Mnemo is a local memory sidecar for developers building custom LLM pipelines, generating context injectable into subsequent prompts through entity relationship extraction, SQLite persistence, and graph expansion retrieval.
Who it is for: Developers building custom LLM pipelines, teams wanting cross-session persistent memory, and teams wishing to control memory data and runtime environment locally. Developers using hosted Agent harnesses with built-in memory capabilities are not its primary target.
Core capabilities
- Entity and Relationship Extraction: After receiving text, it calls the configured LLM to extract entities such as people, tools, locations, and concepts along with their relationships.
- Entity Merging and SQLite Persistence: Deduplicate entities by name and type, merge aliases, and write results to SQLite.
- Six-Stage Graph Expansion Retrieval: Sequentially perform full-text block search, entity name search, knowledge graph BFS expansion, relationship filtering, scoring sorting, and context_prompt assembly.
- Local Multi-Backend Operation: Runs as a single static binary file and can connect to Ollama, OpenAI, Anthropic, or other backends compatible with the OpenAI API.
Pricing
Pricing and pricing changes are pending verification; it cannot be determined from existing materials whether there is a free tier, subscription tiers, or enterprise pricing.
Updated 2026-08-24