ArchonDev

Govern AI Code Your Way
CLI for total control. Drop-in for your existing tools.

Website • Documentation • Download • Changelog

---

What is ArchonDev?

ArchonDev is an AI-powered development governance system that prevents your AI coding assistant from making architectural mistakes, forgetting context, and introducing regressions.

The Problem: AI coding assistants are powerful but unreliable. They hallucinate, ignore your architecture, forget what they learned, and break things they just fixed.

The Solution: ArchonDev adds governance — architectural rules your AI must follow, quality gates before every change, and persistent learning across sessions.

---

Two Ways to Use ArchonDev

Full CLI — Replace Your Tools

AI terminal that governs itself. Total control with architecture enforcement, dependency tracking, and security sentinels.

npm install -g archondev

Then run:

archon

Content-only requests (stories, outlines, lessons, visuals) use lightweight planning to avoid blocking.

Full Documentation →

---

BYOK Key Security

When you use BYOK, your API keys stay on your machine and are never uploaded to ArchonDev servers. The CLI encrypts keys at rest and stores them with owner-only file permissions. BYOK runs locally; cloud execution is Credits-only.

Lite Package — Enhance Your Tools

Governance files for any IDE or AI tool. Drop into Cursor, Claude Code, Windsurf, or any AI coding assistant.

Download for your IDE:

IDE	Download
Cursor	cursor-package.zip (~50KB)
Claude Code / Amp	claude-amp-package.zip (~50KB)
Google Gemini	gemini-package.zip (~50KB)
Windsurf / Codeium	windsurf-package.zip (~50KB)
VS Code + Copilot	vscode-copilot-package.zip (~50KB)
OpenAI Codex	codex-package.zip (~50KB)
Generic (any AI)	generic-package.zip (~50KB)

What's in the package:

File	Purpose
.archon/active/architecture.md	Governance constitution — components, invariants, protected paths, accessibility config
.archon/active/tasks.json	Active task list for multi-item requests
.archon/current_context.md	Current context handoff file for long sessions
AGENTS.md	AI instructions — code review, task extraction, accessibility, SEO, GEO, memory management
DEPENDENCIES.md	Regression prevention — tracks what breaks when you change files
progress.txt	Learning log — persists knowledge across sessions
.archon/config.yaml	Configuration file
archondev-scenarios/	Smart onboarding for new/existing/continuing projects
examples/	Sample workflows and patterns
IDE-specific rules	.cursorrules, CLAUDE.md, GEMINI.md, etc.

Usage:

1. Download the package for your IDE
2. Unzip to your project root
3. Tell your AI: "read .archon/active/architecture.md"

---

VS Code Extension

Real-time violations + quick-fix suggestions. Press Ctrl+. (Cmd+.) for lightbulb menu with auto-fixes.

Download archondev-0.2.0.vsix

---

Features

Core Governance

• Architectural Governance — Define components, boundaries, and invariants your AI must respect
• Quality Level / Posture — Right-sized architecture (prototype/production/enterprise)
• Dependency Tracking — Know what breaks before you change it (DEPENDENCIES.md)
• Learning Persistence — AI remembers patterns across sessions via progress.txt
• Quality Gates — Every change must pass before commit

New in v2.19.6

• Conversational Flow Fixes — Mixed requests like "analyze then help me do X" now continue smoothly into task handling.
• Usage Transparency by Tier — archon usage now reports practical model-level spend:
  • Credits: since-last-top-up context, base model cost, platform fee, and credits deducted.
  • BYOK: estimated provider spend by model for Today, Last 24 Hours, Last 7 Days, Last 30 Days, and Year to Date.
• Usage History Reliability — Fixed profile/user ID resolution in CLI usage commands to prevent false empty usage results.

New in v2.19.30

• Analysis-First by Default — Approval for analysis requests now returns evaluated recommendations first, without creating atoms.
• Explicit Task Conversion — Users can create governed work from analysis with explicit create atom.
• Explicit Execution Control — Chat execution requires explicit execute wording, preventing accidental starts from conversational continue.

New in v2.19.31

• One-Step Plan-First Output — Analysis requests now return recommendation + implementation plan + sample draft in one response.
• Day-1 Scope Accuracy — Day-1 lesson analysis now avoids cross-day contamination (day-10/day-11 files).
• Cleaner Intent Handoff — Users can move from analysis to implementation with natural command save this (or create atom).

New in v2.19.29

• Governance Steering (Not Hijacking) — Architecture/path violations now pause execution with clear corrective guidance instead of hard-failing the run.
• Blocked-State Recovery — Governance constraint hits now move atoms to BLOCKED (actionable correction state) rather than FAILED retry state.
• Approval Context Reliability — approve / approve plan now correctly apply to the pending proposal request in chat mode.

New in v2.19.4

• Design Approval Gate — Plan flow requires explicit design approval before plan generation.
• Plan Structure Enforcement — Plans include explicit verification and acceptance-check steps.
• Root-Cause Bug Intake — Bug reports now capture evidence, reproducibility, and root-cause hypothesis.
• Two-Stage AI Review — Spec compliance review runs before code quality review.

New in v2.19.0

• Governance Store CLI — Status, task updates, handoffs, and legacy migration via archon governance
• AGD Migration — Move legacy governance files into .archon/active|history|archive
• Governance SQLite/FTS — Search architecture, tasks, handoffs, and decisions
• Lite Packages Updated — New AGD layout with task and handoff templates

New in v2.18.0

• Automated Model Registry Sync — Daily pricing verification against provider docs
  • New models: Claude Opus 4.6, GPT-5, GPT-5 Mini
  • Confidence scoring prevents bad parses from corrupting billing
  • CLI: archon models sync, archon models list

New in v2.17.0

• CLI Hardening — Security and governance fixes
  • Supabase SDK no longer imported directly in CLI (uses core helpers)
  • Git commit hardened against shell injection
  • Rollback scoped to atom-touched files only
  • API key entry masked, plan editing supported

New in v2.16.0

• Intent Detection — Smart routing for codebase analysis vs implementation
  • New explore intent routes analysis requests to summary flow
  • Polite prefixes ("Please", "Can you") no longer break pattern matching
  • runExploreFlow() shows project summary without triggering planning

New in v2.15.0

• Credits Usage Dashboard — Balance and model usage stats on startup
  • Color-coded balance display, top 3 model usage breakdown

New in v2.14.0

• Auth & Payment UX Fixes — 5 critical fixes (timeout bug, BYOK flow, payment clarity)

New in v2.12.0

• Billing Audit & Reconciliation — Immutable audit log, daily balance reconciliation, discrepancy alerts
  • CLI: archon credits audit

New in v2.11.0

• Atomic Billing Security — Complete billing rewrite with server-side tier checks, atomic transactions, idempotency keys, race condition prevention

New in v2.10.0

• Multi-Provider Support — OpenAI and Google AI alongside Anthropic
• Database-Driven Model Registry — Pricing stored in Supabase for instant updates

New in v2.6.0

• Smart Model Routing — Automatic cost optimization for AI operations
  • Three model tiers: PLANNING ($$$), REASONING ($$), EXECUTION ($)
  • Operations automatically routed to optimal tier
  • Configure per-tier model preferences via archon preferences
  • 80-95% cost savings on routine tasks

New in v2.5.0

• Streamlined Onboarding — Complete rewrite of first-run experience
  • Login required first — Creates user account before any other steps
  • Tier selection upfront — Choose FREE/BYOK/Credits before AI calls
  • Fixed double-init bug that confused new users
  • Clear, sequential output — no more jumbled messages
• AI-Powered Conversational Interview — Natural conversation replaces rigid menus
  • AI asks follow-up questions based on your answers
  • Infers project details from natural language
  • Uses Haiku model for cost-efficient onboarding
  • Falls back to simple prompts when no API key available
  • Say "skip" or "just start" anytime to use defaults
• Update Notifications — Automatic version checking on startup
  • Background check against npm registry (non-blocking)
  • Shows banner when new version available
  • Displays update command: npm update -g archondev

New in v2.4.0

• Dependency Graph Scheduler — Analyze atom dependencies for parallel execution
  • Wave-based scheduling using topological sort
  • CLI: archon parallel schedule, archon parallel run-waves
• Eject Command — Clean removal of ArchonDev from any project
  • CLI: archon eject --dry-run
• Atomic Rollback — Revert individual atom commits via git
  • CLI: archon revert <atom-id>, archon history

New in v2.3.0

• 5-Phase Interview System — Conversational interview to define your project
  • Phases: Discovery → Features → Technical → Branding → Review
  • CLI: archon interview, archon interview show, archon interview validate
• Project Constitution — Immutable specification generated from interview
  • Formal contract between you and AI for what gets built
  • Complexity tier calculation (SIMPLE/MODERATE/COMPLEX/ENTERPRISE)
  • Build hours and cost estimation
  • SHA-256 hash for integrity verification
• Challenge Mode — AI pushback on scope creep during review
  • Detects too many features, high complexity, vague requirements
  • Suggests features to defer to post-MVP
  • Scope score (0-100) to identify risk
• Atom Generator — Transform Constitution into prd.json
  • CLI: archon generate, archon generate --dry-run
  • Auto-generates setup, database, auth atoms
  • Splits complex features into backend/frontend atoms

New in v2.1.0

• GitHub Integration — Connect GitHub for cloud execution with automatic PR creation
  • Connect: archon github connect
  • Status: archon github status
  • Cloud agents clone repos, make changes, and submit PRs automatically
• Cloud Semantic Indexing — Semantic search stored in Supabase pgvector
  • Setup: archon index init --cloud
  • Index: archon index update --cloud
  • Search: archon index search "query" --cloud
  • Cost: ~$0.01-0.30/month storage, queries are free
• Webhook-Triggered Worker — Cloud worker with instant job pickup and auto-stop for cost savings

New in v2.0.0

• Cloud Agents — Run agents in the cloud, close your laptop, check progress anywhere
  • Execute: archon execute ATOM-001 --cloud
  • Watch: archon execute ATOM-001 --cloud --watch
  • Status: archon cloud status
  • Requires: archon github connect first
• Cross-Device Sessions — Save session, continue on another machine
  • Save: archon session save
  • Resume: archon session resume [id]
• Parallel Agents — Run multiple agents simultaneously with git worktrees
  • Execute: archon execute --parallel ATOM-001 ATOM-002
  • Manage: archon parallel status, archon parallel merge
• One-Click Deploy — Auto-detect platform and deploy
  • Deploy: archon deploy
  • Preview: archon deploy --preview
• Semantic Indexing — AI-powered codebase search (local with Ollama or cloud with pgvector)
  • Local: archon index init --local
  • Cloud: archon index init --cloud
  • Search: archon index search "query"
• Smart Orchestration — System suggests optimal execution mode based on project size
  • Configure: archon preferences execution-set parallel.mode always

New in v1.9.0

• Quality Level / Posture — Tell AI how rigorous to be
  • prototype: Fast iteration, minimal governance, skip complex patterns
  • production: Secure defaults, basic monitoring, modular design (default)
  • enterprise: Full governance with audit logging, RBAC, SLOs, compliance
  • Applies to architecture, code generation, AND code review
  • Prevents over-engineering for simple projects
• Trigger: Set qualityLevel.posture in .archon/active/architecture.md

New in v1.8.0

• SEO Optimization — Automated meta tags, Open Graph, Twitter Cards
  • AI scans, identifies gaps, generates missing tags
  • User approves before changes are applied
  • Trigger: seo check, seo fix, add open graph
• GEO for AI Search — Optimize for ChatGPT, Perplexity, Claude citations
  • AI generates 3 candidate 7-word brand phrases
  • AI generates 3 candidate 50-word descriptions
  • User selects preferred identity
  • JSON-LD schemas for AI comprehension
  • Trigger: geo identity, geo schema

New in v1.7.0

• Pre-Deploy Accessibility Check — WCAG 2.2 AA compliance before going live
  • Legal liability warnings (ADA, EAA, Section 508)
  • Auto-fix for common issues (contrast, alt text, focus)
  • WCAG 2.2 AA badge for compliant sites
  • CLI: archon a11y check, archon a11y fix, archon a11y badge

New in v1.6.x

• Task Extraction Protocol — AI confirms all items in multi-item requests before starting
  • Prevents lost requirements (AI often forgets items 3+ in a list)
  • Trigger phrases: plan these tasks, task status, what's on my list
• Smart Onboarding — Detects new project, existing project, or continuing session
• Code Review Mode — AI reviews code without modifying it

Tools & Extensions

• Local Database — Optional SQLite for tracking atoms and learnings
• VS Code Extension — Real-time diagnostics with quick-fix suggestions

---

CLI Commands

Command	Description
archon	Interactive mode
archon interview	5-phase project interview
archon generate	Generate prd.json from Constitution
archon plan <description>	Create governed work item
archon execute <atom-id>	Execute with quality gates
archon execute --cloud	Execute in cloud (creates PR)
archon execute --parallel	Execute multiple atoms in parallel
archon github connect	Link GitHub account for cloud execution
archon github status	Check GitHub connection status
archon cloud status	List cloud executions
archon session save	Save session for cross-device
archon session resume	Resume saved session
archon parallel status	Show parallel execution status
archon deploy	One-click deploy (auto-detect platform)
archon index init [--local|--cloud]	Initialize semantic index
archon index update [--cloud]	Index changed files
archon index search <query> [--cloud]	Semantic codebase search
archon review init	Initialize code review
archon a11y check	Run WCAG 2.2 AA audit
archon deps list	View dependency rules
archon watch	Live TUI dashboard
archon seo check	Run SEO meta tag audit
archon geo identity	Generate brand identity phrases

Full CLI Reference →

---

Pricing

Tier	Cost	What You Get
Free	$0	Ultra-cheap models (GPT-5-nano, Gemini Flash-Lite)
Credits	Pay as you go	All models, 10% service fee
BYOK	$0	Use your own API keys

No subscriptions. No commitments.

---

Links

• Website: archondev.io
• Documentation: archondev.io/docs
• Changelog: archondev.io/changelog
• Download Lite Packages: archondev.io/download

---

License

MIT © Jumping Ahead Corp.

---

_{Built with the Ralph Protocol — fresh context, persistent learning.}