{"id":"rockywuest-qdrant-mcp-pi5","name":"qdrant-mcp-pi5","af_score":57.8,"security_score":29.0,"reliability_score":26.2,"what_it_does":"Provides a Model Context Protocol (MCP) server integration for persistent semantic memory on Raspberry Pi 5 using local Qdrant storage. Agents can store and retrieve facts via semantic similarity (embeddings) through mcporter (MCP client/bridge), optionally with an OpenClaw plugin that hard-enforces memory recall before responses.","best_when":"Running on-device (Raspberry Pi) or other local environments where data must stay local and a simple MCP-tool bridge workflow is acceptable.","avoid_when":"When you need a long-running, network-exposed API with robust built-in auth/observability, or when you cannot manage embedding model downloads/updates and local storage configuration.","last_evaluated":"2026-03-30T15:42:38.312569+00:00","has_mcp":true,"has_api":false,"auth_methods":["Local process/STDIO MCP via mcporter (no explicit auth described)","Optional embedding provider access for Gemini embeddings (not fully specified)"],"has_free_tier":false,"known_gotchas":["First call slower due to embedding model load; agents may time out if they expect warm calls","Environment variable expansion caveat for tilde (~) in mcporter env values","Do not use QDRANT_URL and QDRANT_LOCAL_PATH simultaneously (configuration conflict)","OpenClaw gateway config may not support mcpServers; README suggests using mcporter bridge instead"],"error_quality":0.0}