Sovereign Intelligence

PurpleDirective

A self-improving multi-agent AI system built for one human operator. Hierarchical orchestration, structured deliberation, dual-runtime deployment, and closed-loop learning.

Human (Final Authority) | +-- Violet (COO — Triage, Delegation, Audit, Synthesis) | +-- E.I.K. Advisory Council (2-round deliberation) | +-- Evolution — forward motion, innovation | +-- Improvement — verification, quality assurance | +-- Keenness — edge-watching, blind spots | +-- Operational Staff (domain-specific execution) +-- Domain agents scaffolded from template +-- Each with: directives, tools, identity, data

How It Works

Triage

Every task is classified before any agent touches it. Simple tasks go to operational staff. Complex questions route through the advisory council first.

Routing

Deliberation

Three advisory agents with structurally differentiated perspectives analyze complex questions in parallel. Round 1: independent analysis. Round 2: cross-response. Hard stop.

E.I.K. Protocol

Execution

Operational agents follow the DOE framework — Directive (what to do), Orchestration (how to decide), Execution (deterministic scripts). Every output is audited.

DOE Framework

Self-Improvement

Cloud sessions compile knowledge into local memory. Errors are corrected and logged. The governing protocols evolve. The system gets measurably smarter with every session.

Closed Loop

Dual Runtime

The same agents run on both cloud and local inference. Knowledge transfers between runtimes automatically.

Cloud

Claude via Claude Code. Full identity profiles, multi-agent deliberation, complex analysis. The power of frontier models with structured orchestration.

Claude Code

Local

Ollama via Purple-Directive: CLI. Condensed identities under 2K tokens, MCP tool support, offline-capable. Sovereign inference on your own hardware.

Purple CLI

Open Source

The framework, protocols, and tools are public. Build your own agent organization.

Violet

The COO agent framework. Advisory council, operational agents, memory kernel, teaching protocol, audit system.

View on GitHub →

Purple CLI

Local AI runtime. Python application connecting to Ollama with MCP tools, agent identities, and plan/build modes.

View on GitHub →