nfemmanuel/iranti
🔌 MCP Servernfemmanuel
Persistent shared memory layer for AI coding agents to maintain context across sessions and tools.
Iranti addresses the 'short-term memory' limitation of current AI coding agents by introducing a structured, persistent memory layer. Built as an MCP server, it acts as a centralized knowledge base that agents can query and update in real-time. The system utilizes an entity/key/value triple architecture, which allows for highly granular data storage and retrieval.
Key technical features include hybrid semantic search, which combines vector-based similarity with exact keyword matching to ensure agents retrieve the most relevant context. It also implements robust conflict resolution mechanisms to handle concurrent updates from multiple agents or sessions. By providing a standardized interface for state management, Iranti enables seamless hand-offs between different coding tools, ensuring that task checkpoints and project-specific configurations remain accessible throughout the entire development lifecycle. This infrastructure is essential for building autonomous agents that require deep, multi-session project awareness.
💡Highlights
- ├─Entity/key/value triple storage
- ├─Hybrid semantic search engine
- └─Cross-agent state synchronization
🎯For
- ├─AI Software Engineers
- └─Agentic Workflow Developers