Modules
Understanding the NarraNexus module system and how capabilities are organized
What Are Modules?
Modules are the primary unit of capability in NarraNexus. Each module encapsulates a specific domain of functionality -- such as memory management, conversation handling, or background job scheduling -- and integrates with the Agent Runtime through a standardized interface. The platform ships with nine built-in modules that cover the core needs of AI agent operation.
Module Interface
Every module implements a consistent interface that allows it to participate in the agent runtime pipeline. Modules register hooks that are called during specific pipeline steps, and they expose MCP tools that the LLM can invoke during the Tool Execution step. This dual integration means modules can both passively influence agent behavior (via hooks) and actively provide capabilities (via tools).
Built-in Modules
NarraNexus includes the following modules:
| Module | Purpose |
|---|---|
| Memory | Long-term storage and retrieval via EverMemOS |
| Awareness | Context extraction and entity recognition |
| Chat | Conversation history and session management |
| Social Network | Entity graphs and relationship tracking |
| Jobs | Scheduled, periodic, and continuous task execution |
| RAG | Document retrieval via Google Gemini File Search |
| Skills | ClawHub marketplace for shareable capabilities |
| Matrix | Inter-agent communication via Synapse |
| Event Memory | Episodic storage for significant events |
Creating Custom Modules
The module system is extensible. Custom modules follow the same interface pattern: define hooks for runtime integration and register MCP tools using the fastmcp library. This allows teams to add domain-specific capabilities without modifying the core platform.