ArcKit: The Enterprise Architecture Governance Harness

Build better enterprise architecture through structured strategy, design, delivery, and assurance workflows.

ArcKit is a toolkit for enterprise architects that transforms architecture governance from scattered documents into a systematic, AI-assisted workflow for principles, stakeholder analysis, risk management, business cases, requirements, data modeling, technology research, strategic planning, vendor procurement, design reviews, and compliance.

Quick Start

Installation

Claude Code (premier experience — install plugin, requires v2.1.172+):

claude install latest
# In Claude Code:
/plugin marketplace add tractorjuice/arckit-claude

Install overlays as needed:

# Core (75 commands — UK Government + generic enterprise)
claude plugin install arckit@arckit-claude
# Regional + sector overlays
claude plugin install arckit arckit-{uae,fr,ca,eu,at,au,us,uk-nhs,uk-gcloud}
# Enterprise architecture + AI agent governance
claude plugin install arckit arckit-togaf-adm arckit-agent-architecture

CLI (Copilot, Codex, OpenCode):

pipx install arckit-cli  # or: uv tool install arckit-cli
arckit init my-project --ai copilot  # or --ai codex / --ai opencode

Gemini CLI:

gemini extensions install https://github.com/tractorjuice/arckit-gemini

One-liner (auto-detects pipx → uv → pip --user):

curl -fsSL https://raw.githubusercontent.com/terrygzhou/arc-kit/main/install.sh | bash

Latest Release: v6.2.1

Initialize a Project

# Claude Code / Gemini: no init needed — plugins/extensions provide everything
arckit init my-project --ai copilot  # or --ai codex / --ai opencode
cd my-project && code .  # then use /arckit-* commands in Copilot Chat

Upgrading

# CLI: upgrade tool + re-init project
pipx upgrade arckit-cli
cd /path/to/project && arckit init --here --ai copilot
# Claude Code: automatic via marketplace
# Gemini: gemini extensions update arckit

CLI Build

Run a full recipe-driven ADM cycle from the terminal:

arckit build my-project --recipe togaf-adm-full

Build Configuration

Provide structured input via build-config.yaml for deterministic, repeatable builds:

# Explicit path
arckit build my-project --config .arckit/build-config.yaml

# Auto-discovered (tries .arckit/build-config.yaml, then build-config.yaml)
arckit build my-project

Example build-config.yaml (full template at scripts/build-config.example.yaml):

project:
  id: "ent-mod"          # {P} — short ID for artifact naming
  name: "Enterprise Modernization"  # {NAME} — display name

discovery:
  systems:
    - "CRM: Salesforce org, 12 integrations"
    - "ERP: SAP S/4HANA, 8 custom modules"
  capabilities:
    - domain: "Customer Engagement"
      maturity: 3
      notes: "Fragmented across 3 teams"
  pain_points:
    - type: "operational"
      description: "Monthly reconciliation takes 3 weeks"
      impact: "financial"
  constraints:
    - "Budget cap: $2.5M Phase 1"
    - "SOX compliance mandatory"

requirements:
  focus_areas:
    - "cloud migration"
    - "PCI-DSS compliance"
    - "data platform modernization"

stakeholders:
  priorities:
    - role: "CFO"
      priority: "Reduce OpEx by 25% within 18 months"
      weight: high
    - role: "CTO"
      priority: "Eliminate COBOL dependency"
      weight: high
  groups:
    - name: "Business"
      members: ["CFO", "VP Sales", "Head of Marketing"]
    - name: "Technology"
      members: ["CTO", "CISO", "VP Engineering"]

# Optional: override per-phase project IDs (default: auto-derives from {P})
phase_ids:
  DISC: "ent-mod-disc"

Sections — project (always set), discovery ({DISC_SCOPE}), requirements ({REQ_SCOPE}), stakeholders ({STKE_SCOPE}), phase_ids (per-phase {P_<ID>} overrides). All sections optional; missing sections fall back to interactive wave prompts. Config is persisted in build state for --resume.

Subsequent runs resume from the last wave:

arckit build my-project --recipe togaf-adm-full --resume

Key flags:

Flag	Purpose
--recipe <name>	Recipe name or YAML path (default: togaf-adm-full)
--config <path>	Build config YAML with structured placeholder values
--plan	Dry run — print wave plan, do not execute
--resume	Resume from last incomplete wave
--target <ID>	Build only this target and its dependencies
--refresh <ID>	Force-rebuild this target and downstream
--parallel N	Max concurrent LLM calls per wave (default: 4)
--no-commit	Skip per-wave git commits
--base-url URL	Override LLM base URL
--model NAME	Override LLM model

Placeholder inheritance: {P} is captured once at the ADMP wave; each phase gets {P_<ID>} auto-derived (e.g. {P_BPCM} → "{P}-BPCM") and can be overridden independently via config phase_ids.

---

Platform Support

Platform	Claude Code	Gemini CLI	Copilot	Codex/OpenCode	Vibe
macOS	✅	✅	✅	✅	✅
Linux	✅	✅	✅	✅	✅
Windows (WSL2)	✅	✅	✅	✅	✅
Windows (native)	✅	✅	✅	Partial	✅

---

OKF Interoperability

ArcKit can exchange Markdown knowledge bundles using an Open Knowledge Format layer:

• /arckit:export-okf — copy ARC artifacts to OKF bundle with metadata
• /arckit:import-okf — import OKF bundle, materialise as RSCH review notes
• Enable source frontmatter: ARCKIT_OKF_FRONTMATTER=1

---

The ArcKit Workflow

ArcKit guides you through the enterprise architecture lifecycle:

Phase 0-4: Foundation & Strategy

Command	Purpose
/arckit:plan	Project plan with timeline, phases, gates, Mermaid diagrams
/arckit:principles	Enterprise architecture principles (cloud, security, tech standards)
/arckit:stakeholders	Stakeholder drivers, goals, and measurable outcomes
/arckit:risk	Risk register (HM Treasury Orange Book) — 6 categories, 4Ts response
/arckit:sobc	Strategic Outline Business Case (Green Book 5-case model)

Phase 5: Requirements & Data

Command	Purpose
/arckit:requirements	Comprehensive requirements with acceptance criteria
/arckit:platform-design	Multi-sided platform strategy (8 canvases)
/arckit:data-model	Data model with ERD, GDPR compliance, data governance
/arckit:dpia	Data Protection Impact Assessment (UK GDPR Article 35)
/arckit:datascout	External data source discovery and evaluation

Phase 6-7: Research & Strategy

Command	Purpose
/arckit:research	Technology research with build vs buy analysis
/arckit:grants	UK government grants, funding, and accelerator programmes
/arckit:wardley	Strategic Wardley Maps for architecture decisions
/arckit:roadmap	Multi-year architecture roadmap with governance
/arckit:strategy	Executive-level Architecture Strategy synthesis
/arckit:adr	Architecture Decision Records (MADR v4.0)

Phase 8-10: Procurement, Design, Delivery

Command	Purpose
/arckit:sow	Statement of Work / RFP document
/arckit:dos	Digital Outcomes and Specialists procurement 🇬🇧
/arckit:gcloud-search	G-Cloud service search with live marketplace 🇬🇧
/arckit:gcloud-clarify	G-Cloud gap analysis and supplier clarification 🇬🇧
/arckit:evaluate	Vendor evaluation framework and scoring
/arckit:hld-review	High-Level Design review
/arckit:dld-review	Detailed Design review
/arckit:backlog	Prioritised product backlog with sprint planning
/arckit:trello	Export backlog to Trello
/arckit:servicenow	ServiceNow service management design

Phase 12-16: Traceability, Quality, Publishing

Command	Purpose
/arckit:traceability	Requirements traceability matrix
/arckit:analyze	Comprehensive governance quality analysis
/arckit:story	Project story with timeline and governance achievements
/arckit:presentation	MARP slide deck from project artifacts
/arckit:pages	Documentation site with Mermaid rendering

---

Plugin Overlays

TOGAF ADM (arckit-togaf-adm) [COMMUNITY]

Enterprise Architecture Development Method — 12 commands (10 required + 2 optional) covering the full ADM cycle:

Command	Phase	Description
/arckit:discovery	DISC	Current-state baseline (business context, capabilities, applications, data, technology, constraints)
/arckit:adm-preliminary	Preliminary	Architecture vision, scope, drivers
/arckit:business-capability-map	Phase A	Business capability hierarchy
/arckit:application-inventory	Phase C.2	Application catalog with strategic fit
/arckit:data-architecture	Phase C.1	Data entities, governance, reference/master data
/arckit:technology-architecture	Phase D	Technology stack, platforms, infrastructure
/arckit:application-rationalization	Phase C	Keep/merge/replace/retire decisions
/arckit:gap-analysis	Phase E	Current vs target gap matrix
/arckit:transition-architecture	Phase F	Work packages, migration waves
/arckit:architecture-board	Phase G	Board charter, compliance scorecard
/arckit:architecture-change	Phase H	Change requests, ADM re-entry (optional)
/arckit:architecture-repository	Repository	Patterns, standards, reference architectures (optional)

Install: claude plugin install arckit arckit-togaf-adm

AI Agent Architecture (arckit-agent-architecture) [COMMUNITY]

6 commands for autonomous AI agent governance, design, and security:

Command	Description
/arckit:agent-inventory	Agent catalog with capabilities, security classification
/arckit:agent-design	Agent architecture spec — patterns, tools, memory
/arckit:agent-governance	Oversight models, approval workflows, audit
/arckit:agent-integration	Multi-agent orchestration, contracts
/arckit:agent-security	Sandboxing, permissions, injection defences
/arckit:agent-maturity	5×5 maturity model for agent programs

Install: claude plugin install arckit arckit-agent-architecture Combined recipe: claude agent recipes/togaf-agent-full.yaml

UAE Federal (arckit-uae)

AI governance (Charter, Autonomy Tiers), procurement (Federal Decree-Law 11/2023), and cloud residency. Commands chain from principles through procurement.

G-Cloud Bid Authoring (arckit-uk-gcloud) [PROPRIETARY]

10 commands for UK Government supplier bid authoring: supplier profile, service design, SDD lots, declaration, pricing, security, competitor benchmark, review, and submission pack. Install: claude plugin install arckit arckit-uk-gcloud

EU & French Government (arckit-fr, arckit-eu)

17 commands covering GDPR, EU AI Act, NIS2, DORA, CRA, DSA, Data Act, ANSSI hygiene, SecNumCloud, EBIOS Risk Manager, French public procurement, and algorithm transparency.

---

Why ArcKit?

Problem: Traditional enterprise architecture suffers from scattered documents, inconsistent governance, manual vendor evaluation, lost traceability, and stale documentation.

Solution: Template-driven quality, systematic workflows, AI assistance, enforced traceability, and Git-based version control.

---

Supported AI Assistants

Assistant	Support	Notes
Claude Code	✅ Premier	Primary platform — plugin with agents, hooks, MCP servers
Gemini CLI	✅ Full	Extension with commands and MCP servers
GitHub Copilot	✅ Core	Prompt files, custom agents, repo-wide instructions
OpenAI Codex CLI	✅ Core	CLI with commands and templates
OpenCode CLI	✅ Core	CLI with commands and templates

Claude Code provides unique capabilities: parallel /arckit:build harness, autonomous research agents (10 agents), session hooks (auto-detect, context injection, filename correction, MCP auto-allow), and per-command stop hooks (output validation).

---

Project Structure

my-project/
├── .arckit/
│   ├── scripts/          # Automation scripts
│   ├── templates/        # Default templates (refreshed by arckit init)
│   └── templates-custom/ # Your customizations (preserved)
├── projects/
│   ├── 000-global/       # Global principles
│   └── 001-project/      # Project artifacts (ARC-NNN-TYPE-vN.N.md)
├── .agents/skills/       # Codex CLI skills
├── .codex/               # Codex agents + config
├── .github/              # Copilot prompts + agents
└── .opencode/            # OpenCode commands

---

Plugin Footprint

• Always-on per session: ~10,042 tokens (73 command-skills + 5 utility skills + 16 agents)
• On-invoke: ~250 to ~60K tokens per command (most 5–10K range)
• Utility skills use paths: globs to scope always-on cost to relevant projects
• Community overlays add their own always-on baseline — install only what you need

---

Token Limit Troubleshooting

If you see: API Error: Claude's response exceeded the 32000 output token maximum

• Team/Enterprise plans: export CLAUDE_CODE_MAX_OUTPUT_TOKENS=64000
• All plans: Use Write tool strategy — /arckit:requirements but write directly to file using Write tool, show me only a summary
• High-risk commands: /arckit:sobc, /arckit:requirements, /arckit:data-model, /arckit:sow
• Research commands run as autonomous agents (separate context windows)

See full guide: docs/TOKEN-LIMITS.md

---

Documentation

• Lifecycle visuals: docs/WORKFLOW-DIAGRAMS.md, docs/DEPENDENCY-MATRIX.md
• Core guides: docs/guides/principles.md, docs/guides/requirements.md, docs/guides/procurement.md, docs/guides/design-review.md
• Traceability: docs/guides/traceability.md
• DDaT Role Guides: docs/guides/roles/ — 18 guides mapping commands to UK DDaT roles

---

Contributing

We welcome contributions! See CONTRIBUTING.md.

Priority areas: Enterprise tool integrations (Jira, Azure DevOps), additional AI agent support, template improvements, documentation, ServiceNow API integration.

---

Support & License

• Issues: GitHub Issues
• Releases: GitHub Releases
• Latest: v6.2.1

MIT License — see LICENSE.

Exception: plugins/arckit-uk-gcloud/ is proprietary (not MIT).

Built with ❤️ for enterprise architects who want systematic, AI-assisted governance.