Why DevArch?
The hard part of software isn't writing code — it's keeping the codebase coherent as it grows. DevArch enforces the practices that matter: planning before implementation, tests for every command, session context that survives between conversations, and boundaries that don't leak.
For complex domains, it goes further. Describe your business in conversation — the tooling captures bounded contexts, domain events, and ubiquitous language behind the scenes. The model constrains the code, not the other way around.
Skill libraries like claude-code-best-practices and agent-skills are valuable — and complementary. They give an AI agent good habits for a single session. DevArch gives a project continuity across sessions: planning that decomposes into phases, context that survives between conversations, agents that audit and checkpoint your work, and a capability system that adapts as your domain grows. The skill libraries make each session better. DevArch makes the project better.
By the Numbers
Audited across 5 active repositories from January through April 2026 — one developer, real production code. Over 97 days, DevArch helped produce 522,642 lines of tested code across 7 languages, backed by 4,046 unit tests and 1,025 commits (90% AI-assisted). At the COCOMO II industry average of 15 debugged, tested LOC per developer per day, that output represents roughly 158 man-years of traditional development. See the workflow →
DevArch in Action
Real output from production sessions — init, audits, planning, implementation, and commit. All autonomous.
CLI Walkthrough
Real terminal output from production sessions — init, audits, planning, implementation, and commit. All autonomous, nothing staged.
View walkthrough →Generated Code
Domain models, bounded contexts, event-driven handlers, and full test suites — all produced by DevArch-guided sessions.
View examples →What DevArch Does
Methodology + notation + agents. Conversation is the interface.
Methodology
DDD principles adapted for AI-assisted workflows. Start with the base. Activate domain modeling only when complexity demands it.
Notation
A structured format that gives your conversations persistent state. You never touch it — the agents read and write it behind the scenes.
Agents
Specialized behaviors for Claude Code: session audits, mutation verification, domain discovery, code scaffolding. They run automatically.
Conversation
You describe your domain. The tooling listens and produces structured output. "Show me the model" gives you a summary, not a YAML dump.
Behavior Statements
Every side-effect function gets a structured statement — DOES, WHEN, BECAUSE, REJECTS WHEN — before any test is written. If you can't state what the code does, it's not ready to test.
Test Grading
Tests are derived directly from Behavior Statements and graded RED/YELLOW/GREEN. Only GREEN ships — no tautologies, no mock-only assertions, no missing state checks.
Architecture Decision Records
When a decision constrains future sessions, the agent asks "ADR-worthy?" Approved decisions are promoted from session notes to permanent record with context, rationale, and consequences.
Architect Review
On-demand architectural review across 13 categories — requirements, complexity, security, dependencies, cost, and more. The project profile filters out what's irrelevant, so a CLI tool skips cloud cost and a library skips observability.
How It Works
Install once. Init each project. Then just work — the loop handles itself.
Install
One command sets up agents, hooks, and skills in your Claude Code environment.
Init
Initialize a project with directory structure, CLAUDE.md, and coding discipline.
Plan
Describe your goal. The session planner decomposes it into phases.
Finalize
Work summary, commit, and push — all in one command.
Recap
Previous session context is restored automatically at startup.
Plan
Resume the current phase or plan the next one.
Finalize
Summarize, commit, push. Pick up next session where you left off.
Pricing
One product, everything included. Annual license with repo access, updates, and all future agents.
One Developer
Full access to the DevArch repo — all agents, hooks, notation schema, methodology docs, skills, and templates. One seat.
Up to 10 Developers
Everything in Solo for your whole team. Ten seats, one license.
Unlimited Developers
Unlimited seats for your organization. The same repo access, the same tooling — no per-developer math.
All plans are annual. Licensed users get read access to the private GitHub repo — pull updates anytime. No runtime dependencies, no API keys, no usage metering. Use the contact form below to get started.
License Inquiries
Interested in DevArch for your team? Want to discuss licensing or have questions? Drop us a line.