Get Running in Minutes

No LLM needed to start. Full pipeline runs with mock provider.

terminal
$ pip install k9-aif
Successfully installed k9-aif
$ k9aif init
Scaffolding K9-AIF project structure...
$ python main.py
Squad initialized. Agents registered. Pipeline running.
$ k9aif doctor
✓ All checks passed
$ k9aif --help
Usage: k9aif [OPTIONS] COMMAND [ARGS]...

Architecture Hierarchy

A deliberate, layered execution model — every component has a defined role.

Event
Entry point
Router
Routes by intent
Orchestrator
Coordinates domain
Squad
Executes flow
Agent
Implements execute()
LLM
Governed inference
Governance Model Router Zero Trust Security Observability Persistence Messaging Inference

Applying architecture principles, patterns, and governance to the design and execution of agentic process flows and integrations.

Web Apps
APIs
CrewAI
LangChain
IBM Watsonx
BPMN / Blueworks
External Agents
K9-AIF ARCHITECTURE BUS
EVENT DRIVEN  ·  GOVERNED  ·  OBSERVABLE  ·  SECURE
Router
Orchestrator
Squad
Agent
Governance
Zero Trust
Model Router
Inference
Observability
Persistence
Messaging
Graph
Storage
Security

Framework Principles

Architecture First

Design the system before writing code. ABBs define contracts. SBBs implement them.

Governed Execution

Every agent boundary enforces governance. Policy-checked before action executes.

ABB / SBB Separation

Stable abstract contracts. Swappable concrete implementations. Eclipse-style extension points.

Zero Trust Execution

Every agent action verified and risk-evaluated prior to execution at the architecture layer.

Pluggable Providers

Ollama, Watsonx, OpenAI — swap via config. No code changes required.

Observable by Default

Every routing decision persisted. Full audit trail without extra instrumentation.

Enterprise Integration

Kafka, PostgreSQL, Neo4j, MCP. Production infrastructure, not demo infrastructure.

Vendor Neutral

No lock-in. LLM provider, messaging, persistence — all configurable, all replaceable.

CLI Experience

K9-AIF ships with a full command-line interface. Verify, inspect, generate, and run — from the terminal.

k9aif

Ready to build in VS Code + Claude Code

K9-AIF is available as an installable Python package. Open a terminal, install the library, run k9aif init and start building governed multi-agent systems — right inside your IDE.

pip install k9-aif k9aif init python main.py

Where K9-AIF Fits

K9-AIF defines how the full agentic AI system is architected.

  • Frameworks like CrewAI define how agents collaborate
  • Cloud platforms like Azure and AWS provide infrastructure
  • Runtimes execute workflows

About K9-AIF

K9-AIF did not begin as a grand architectural vision. It began as a retrieval capability — and evolved into an architecture-first framework for governed agentic AI systems.

Read how K9-AIF started →

The architecture that outlasts the technology →

Ravi Natarajan
AI Systems Architect — Agentic AI · Multi-Agent Systems · Enterprise Architecture