N3RV
INVISIBLE ENGINEERING INFRASTRUCTURE FOR AI AGENTS
"Because an angel without a harness is unmanageable."
Core Architecture
MAGI Memory
Dual-layer memory routing with ChromaDB for semantic long-term retrieval and SQLite for ACID-compliant session state. Persistent context across sessions.
A2A Hub
Agent-to-Agent protocol for task delegation. Sub-agents communicate via a structured hub — isolated by responsibility, coordinated by design.
SDD Workflow
Spec-Driven Development — 8 phases from explore to archive. Each phase has dedicated agents, skills, and quality gates.
MCP Protocol
4 native MCP servers — memory (12), hub (5), exec (8, 20+ langs, XXH3 cache), code-graph (6, AST+FTS5) — 31 tools total.
18 Agent Skills
Specialized skill files for code, testing, commits, GitHub ops, project-board, SDD phases (8) + Spec-Kit (3), and adversarial review. Loaded contextually per task.
LCL Injection
Automatic parsing of pyproject.toml, package.json manifests. The agent "breathes" your project architecture from session one.
Quick Start
# Install N3RV uv tool install git+https://github.com/juanmanueldaza/n3rv.git # Initialize in your project cd your-project n3rv init