mcp-memory-service v10.60 — Now Available
Memory for AI Agents
Persistent memory for AI agents — REST API, MCP, OAuth, CLI, dashboard. Semantic search, knowledge graph, automatic consolidation. One self-hosted service, every transport.
mcp-memory-service v10.60 — Now Available
Persistent memory for AI agents — REST API, MCP, OAuth, CLI, dashboard. Semantic search, knowledge graph, automatic consolidation. One self-hosted service, every transport.
Temporal contradiction detection, Milvus instance-level graph cache fix, and Kiro CLI harvest support. ~2,005 tests.
New opt-in module detects contradicting facts using an embedding similarity band (0.4–0.75). Emits CONTRADICTED_BY graph edges and marks the older memory with superseded_by. Enable via MCP_CONTRADICTION_DETECTION_ENABLED=true.
Replaces the class-variable graph storage cache with an instance attribute + double-checked locking, preventing cross-instance contamination in multi-test environments. retrieve() now filters superseded_by memories to match sqlite_vec behavior.
The TranscriptParser now auto-detects Kiro CLI session format (kind key) in addition to Claude Code format (type key), mapping kind: Prompt → user and kind: Response → assistant. Fully additive, no breaking changes.
Claude.ai connects to your memory server through Streamable HTTP with OAuth 2.1 authentication.
Battle-tested with comprehensive testing and optimized for performance.
Install from PyPI and connect your AI agents to persistent memory.