Get Started
Cognitive Orchestration (CO)
Section titled “Cognitive Orchestration (CO)”CO is a methodology for structured human-AI collaboration. It works in any professional domain. You do not need to write code to use it.
Claude Desktop: Set up CO for your field in 10 minutes. Choose from seven domain applications (coding, research, learning, finance, education, governance, compliance) or build your own from the CO template.
CO runs via Claude Code (CLI, full enforcement) or Claude Desktop Cowork (plugin, broader accessibility). The Claude Desktop guide covers the plugin path.
CC: Claude Code Patterns
Section titled “CC: Claude Code Patterns”CC is the universal set of patterns for structuring AI-assisted work using Claude Code’s artifact system. It defines how to build agents, skills, rules, hooks, commands, and settings that make any Claude Code session disciplined rather than ad-hoc.
Claude Code Patterns: The seven artifact types, quality rules, and key patterns (progressive disclosure, defense in depth, anti-amnesia).
COC: CO for Codegen
Section titled “COC: CO for Codegen”COC is the software development application of CO. It uses Claude Code as the implementation partner: 50+ specialist agents, framework-aware skills, automated code review, test-driven implementation, and a six-phase workflow from idea to production.
Claude Code Quickstart: Clone the template, run claude, type /start.
Open-source software
Section titled “Open-source software”The Terrene Foundation maintains reference implementations of its specifications. All software is licensed under Apache 2.0: use it, extend it, build commercial products on it.
Kailash Python SDK
Section titled “Kailash Python SDK”The workflow orchestration platform with integrated trust infrastructure. 188 workflow nodes, sync and async runtimes, and the Trust Plane built in. 322,000+ lines of source code. 13,301 tests.
- Install:
pip install kailash - Source: github.com/terrene-foundation/kailash-py
Frameworks
Section titled “Frameworks”Primitives (deterministic building blocks)
Section titled “Primitives (deterministic building blocks)”| Package | Version | Purpose | Install |
|---|---|---|---|
| Kailash Core | 2.3.4 | Workflow orchestration, 188 nodes, sync and async runtimes | pip install kailash |
| DataFlow | 1.4.0 | Zero-config database operations (11 nodes per model), Express CRUD (0.27ms) | pip install kailash-dataflow |
| Nexus | 1.6.1 | Multi-channel deployment (API + CLI + MCP simultaneously) | pip install kailash-nexus |
| Kaizen | 2.3.3 | AI agent primitives: agents, signatures, tool registry, orchestration scaffolding | pip install kailash-kaizen |
| PACT | 0.5.0 | Governance primitives: D/T/R addressing, constraint envelopes, cost tracking | pip install kailash-pact |
Engines (composed, deterministic, cross-cutting)
Section titled “Engines (composed, deterministic, cross-cutting)”Each engine wraps its primitive with additional capabilities and reaches across all five primitives. Engines sit above the governance line: they are deterministic and require no LLM.
| Engine | What it composes | Install |
|---|---|---|
| DataFlowEngine | DataFlow + query monitoring + field validation + data classification policies | pip install kailash-dataflow |
| NexusEngine | Nexus + enterprise middleware (CSRF, audit, metrics, rate limiting, security headers) | pip install kailash-nexus |
| PactEngine | GovernanceEngine + CostTracker + EventBus; the Trust Plane facade | pip install kailash-pact |
Agents (autonomous, below the governance line)
Section titled “Agents (autonomous, below the governance line)”Agents operate below the governance line. They use an LLM to reason, plan, and act, but every action passes through the engines above the line before execution proceeds.
| Agent | What it does | Install |
|---|---|---|
| Delegate | Single autonomous agent; TAOD loop (Think, Act, Observe, Decide) with governance gate | pip install kaizen-agents |
| GovernedSupervisor | Multi-agent orchestration; 7 governance subsystems (accountability, budget, clearance, cascade, dereliction, bypass, audit); progressive disclosure API | pip install kaizen-agents |
Trust Plane
Section titled “Trust Plane”The Trust Plane is the cryptographic trust infrastructure integrated into Kailash v2. It is not a separate product; it is the enforcement layer through which every governed action passes.
76,000+ lines. 4,255 tests. 72 security findings resolved across four red team rounds.
What it provides:
- Human origin tracking. Every AI action traces through a cryptographic chain to the human who authorized it.
- Constraint propagation. Constraints flow through delegation chains and can only tighten, never loosen. This is enforced mechanically.
- Ed25519 signing. Decision records are signed, chained, and tamper-evident.
- RFC 3161 timestamping. Audit trails are independently timestamped.
- Five verification levels. From lightweight logging to full cryptographic attestation.
- PACT for MCP. Governance middleware for any MCP-compatible agent (Claude Code, Codex, Gemini).
Progressive infrastructure
Section titled “Progressive infrastructure”Start with zero configuration. Scale by changing environment variables.
| Level | Infrastructure | What changes |
|---|---|---|
| 0 | SQLite (default) | Nothing. Works out of the box. |
| 1 | PostgreSQL or MySQL | Set KAILASH_DATABASE_URL. Same code. |
| 2 | Multi-worker with Redis | Add KAILASH_REDIS_URL. Task queue, worker registry, distributed locking. |
CARE Platform
Section titled “CARE Platform”The governance platform implementing the CARE specification. Provides the Trust Plane and Execution Plane architecture, constraint management, and reflective improvement infrastructure. Governance enforced at execution time, not observed after the fact.
EATP SDK
Section titled “EATP SDK”The trust protocol implementation. Dataclass-based decision records, Ed25519 signing, HMAC integrity verification, monotonic trust escalation, and multiple store backends.
- Source: github.com/terrene-foundation/kailash-py (packages/eatp)
Praxis
Section titled “Praxis”The practical toolkit for applying CO methodology. Templates, patterns, and tooling for teams adopting Cognitive Orchestration.
Contributing
Section titled “Contributing”All Foundation software accepts contributions under the Apache 2.0 license. Contributions are subject to the Foundation’s contributor framework: intellectual property is irrevocably transferred to the Foundation upon contribution, ensuring no single entity holds exclusive rights.
See each repository’s CONTRIBUTING.md for specific guidelines.