Fegis Memory Server

Example Use Cases

• 1

  Define and run structured tools using simple YAML schemas with typed frames and natural-language instructions.

• 2

  Persistently store complete tool-call context (inputs, outputs, params, timestamps) and embeddings in Qdrant for later retrieval.

• 3

  Semantic search over prior tool usages, including similarity search, filters, and direct lookups to inform decisions or continue workflows.

Description

Fegis enables easy authoring of structured tools using natural-language YAML schemas, automatically captures every tool invocation (inputs, outputs, parameters, timestamps, and session context), and stores vector embeddings in a Qdrant collection for semantic retrieval. It includes a SearchMemory tool and integrates with clients like Claude Desktop for immediate use. The project is configured via environment variables and runs alongside a Qdrant instance, making it useful for agents that need persistent, searchable structured memories of their actions.