- Category
- Developer Tools
- Rank
- No. 885Tools index
Previous survey · No. 870 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- tractorjuice
- GitHub
- 2.2k stars
- Latest release
- v6.14.0
- Date
About
An enterprise-architecture governance toolkit that installs into AI coding assistants as 75+ commands, covering architecture principles, stakeholder and risk analysis, business cases, requirements and data modelling, design reviews and compliance assessment. It gates and provenance-stamps what the assistant produces so the resulting artifacts are audit-ready rather than scattered ad-hoc documents. Its frameworks lean on UK public-sector standards including the HM Treasury Orange Book, Green Book business cases and the GDS Service Standard, with overlays for the EU, Canada, Australia and other jurisdictions.
What it does
ArcKit scaffolds an architecture project around connected Markdown artifacts, guided command workflows, research agents, diagrams, and traceability. Teams can move from initial principles and requirements through design, procurement, assurance, operations, and reporting while preserving dependencies between outputs.
Why it's ranked here
ArcKit stands out for unusually broad workflow coverage and explicit sequencing between architecture artifacts. Its strongest case is not generic document generation, but a coherent operating model spanning research, decisions, compliance, procurement, and delivery. The cost is substantial surface area and uneven capability across supported assistants.
What's good
The dependency model distinguishes mandatory, recommended, and optional inputs, making large document sets easier to navigate. Dedicated workflows cover private-sector, UK government, defence, and AI projects. Research integrations use official cloud documentation, while citation markers preserve source quotes. Packaging converts shared material into several assistant-specific formats instead of maintaining unrelated implementations.
Tradeoffs
Claude Code is the primary and most complete platform, so other integrations do not receive identical capabilities. Native Windows support for the Codex and OpenCode route is partial because some commands contain Bash snippets. Several research services require API keys, and one tender service has no formal availability guarantee. The wide command catalogue also demands deliberate workflow selection.
How to use it well
Use ArcKit when architects need repeatable, linked evidence across a long programme, especially in regulated or public-sector work. Start with planning, principles, stakeholders, risk, and requirements, then follow only the relevant design and assurance path. Treat generated material as governed working documentation. It does not replace implementation tooling, cloud services, or human approval authorities.
Technical notes+
The Python package is declared in pyproject.toml, requires Python 3.11 or newer, exposes the arckit console entry point, and builds with Hatchling plus a custom hatch_build.py hook. src/arckit_cli/__init__.py implements the Typer CLI, resolves packaged shared data, checks required scaffold assets, and configures Codex, OpenCode, Copilot, and Kimi targets. package.json supplies Mermaid and AntV tooling for visual output. docs/MCP-CATALOGUE.md documents six MCP servers and their authentication and allowlisting model. docs/DEPENDENCY-MATRIX.md records artifact dependencies, while docs/WORKFLOW-DIAGRAMS.md turns them into guided project paths. Pytest collection is explicitly rooted at tests in pyproject.toml.
Observed
- License
- MIT for the core project; the UK G-Cloud supplier overlay is explicitly proprietary.
- Primary language
- Python 3, requiring Python 3.11 or newer.
- Packaging
- Hatchling builds the arckit-cli package and includes generated assistant extensions as shared data.
- Interfaces
- Python CLI, Claude Code plugins, Gemini extension, Copilot prompts, Codex and OpenCode scaffolds, Mistral Vibe extension, Kimi plugin, and MCP integrations.
- Platform support
- macOS, Linux, and WSL2 are fully supported; native Windows support is partial for Codex and OpenCode.
- Testing structure
- Pytest is configured to collect tests from a dedicated tests directory.
Read from README.md, package.json, pyproject.toml, src/arckit_cli/__init__.py, docs/README.md, docs/RELEASING.md, docs/TEST-REPOS.md, docs/MCP-CATALOGUE.md, docs/INVESTOR-REPORT.md, docs/DEPENDENCY-MATRIX.md, docs/WORKFLOW-DIAGRAMS.md.
Tags
Tech Stack
Media
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.