Skip to content
Terrene Foundation Terrene Foundation Terrene Foundation Terrene Foundation Terrene Foundation

Claude Code

CO (Cognitive Orchestration) applies to any professional domain: research, finance, education, compliance, governance. This quickstart covers COC, the codegen application. If you work in a different field, see the CO domain applications and the CO template for building your own.

The fastest way to build with Terrene Foundation software is COC (Cognitive Orchestration for Codegen), an AI-assisted development methodology that uses Claude Code as your implementation partner. You describe what you want to build. COC researches, plans, implements, tests, and deploys it, with you in the driver’s seat at every decision point.

COC is not a wrapper or a chatbot. It is a full development environment: 30+ specialist agents, framework-aware skills, automated code review, test-driven implementation, and a phased workflow that takes a project from idea to production.

Getting started with COC

Section titled “Getting started with COC”

1. Install Claude Code

Section titled “1. Install Claude Code”

COC runs best with Claude Code, Anthropic’s CLI-based development tool.

Terminal window
npm install -g @anthropic-ai/claude-code

You will need an Anthropic API key or a Claude Pro/Team subscription. See the Claude Code documentation for setup details.

2. Create your project from the COC template

Section titled “2. Create your project from the COC template”

The COC template repository has everything pre-configured: agents, skills, rules, hooks, and the phased workflow. This is the starting point for any new project.

Terminal window
git clone https://github.com/terrene-foundation/kailash-coc-claude-py.git my-project
cd my-project

That is the entire setup. The COC template repository has everything pre-configured to CO specifications: agents (Intent layer), skills (Context layer), rules (soft guardrails), hooks (hard guardrails), and phased commands that implement the structured workflow CO requires. No configuration needed.

You do not need to clone kailash-py (the SDK source). The COC template is your project template; it pulls in the SDK as a dependency when needed.

3. Start your first session

Section titled “3. Start your first session”

Launch Claude Code from the repository root:

Terminal window
claude

Then type /start. COC introduces itself, explains the workflow, and asks what you want to build.

4. The COC workflow

Section titled “4. The COC workflow”

COC uses a phased approach. Each phase produces artifacts, and each transition requires your approval. Nothing happens without your sign-off.

CommandWhat it does
/startOrientation. COC explains how it works and asks about your project.
/analyzeDeep research. COC investigates your idea, identifies failure points, studies existing patterns, and produces a strategy document.
/todosRoadmap. COC breaks the strategy into ordered milestones and tasks. You review, adjust, approve.
/implementBuild. COC works through tasks one at a time, writing code, tests, and documentation. Run this repeatedly.
/redteamAdversarial testing. COC attacks the project from a real user’s perspective, looking for gaps and failures.
/codifyKnowledge capture. COC writes down what it learned so the next session starts where this one left off.
/deployProduction. COC walks through deployment setup, staging verification, and go-live.

Between sessions, run /wrapup to save progress. The next session picks up exactly where you left off.

5. What COC gives you

Section titled “5. What COC gives you”

Specialist agents. COC has 30+ agents organized by domain. When you need database operations, the dataflow-specialist handles it. When you need an API, the nexus-specialist builds it. When you need AI agent logic, the kaizen-specialist takes over. You do not need to know which agent to call. COC routes automatically based on what the task requires.

Framework-first development. Before writing any code from scratch, COC checks whether the Kailash SDK already has a solution. The SDK has 140+ workflow nodes, zero-config database operations, multi-channel API deployment, and a full AI agent framework. COC knows all of it.

Test-driven implementation. Every feature gets tests before implementation. The tdd-implementer agent writes the test, then writes the code to pass it. The testing-specialist manages a three-tier strategy: unit tests, integration tests with real infrastructure, and end-to-end tests.

Automated review. After every change, the intermediate-reviewer agent performs a code review. Before every commit, the security-reviewer agent checks for vulnerabilities, leaked secrets, and injection risks. This happens automatically. You do not need to request it.

Session continuity. COC remembers what happened across sessions. Workspace state, completed tasks, learned patterns, and session notes all persist. A project that takes ten sessions feels like one continuous conversation.

The Kailash SDK ecosystem

Section titled “The Kailash SDK ecosystem”

COC builds with the Kailash SDK stack. These are the frameworks it draws from:

FrameworkWhat it doesInstall
Kailash CoreWorkflow orchestration with 140+ built-in nodes, runtime execution, error recovery, and constraint propagation. The foundation everything else builds on.pip install kailash
DataFlowZero-config database operations. Define a model, get CRUD nodes automatically. No SQL, no ORM configuration, no migrations to write by hand.pip install kailash-dataflow
NexusDeploy any workflow as an API endpoint, a CLI command, and an MCP tool, simultaneously, from a single definition.pip install kailash-nexus
KaizenAI agent framework. Signatures, multi-agent coordination, tool use, and structured output, all built on the Kailash workflow engine.pip install kailash-kaizen

When you use COC, you rarely install these manually. COC manages dependencies, chooses the right framework for each task, and writes the integration code.

Direct SDK usage

Section titled “Direct SDK usage”

If you prefer to use the SDK without COC, it works standalone:

Terminal window
pip install kailash
from kailash import LocalRuntime, WorkflowBuilder


workflow = WorkflowBuilder()
workflow.add_node("NodeType", "my_node", {"param": "value"})


runtime = LocalRuntime()
results, run_id = runtime.execute(workflow.build())

The SDK is fully documented and usable on its own. But COC is the recommended development experience. It knows the SDK deeply, handles boilerplate, catches mistakes early, and moves significantly faster than manual development.

Next steps

Section titled “Next steps”