AI Development Governance — Stop Babysitting Your AI Agent
The complete AI development system. It manages your entire development process so you can focus on the big picture — not constantly correcting your AI.
# macOS/Linux:
npm install -g archondev && archon
# Windows PowerShell:
npm install -g archondev; archonWhat you get:
- AI that reads and respects your architecture before writing code
- Quality Level / Posture — Right-sized architecture (prototype/production/enterprise)
- Adversarial planning (Architect proposes, Sentinel critiques)
- Automatic quality gates before changes are applied
- Learning persistence — mistakes are remembered and avoided
- Dependency tracking — prevent regressions with "what-breaks-what" map
- Bug reporting with root cause analysis
- AI-powered code review for any codebase
- Multi-provider key support with adversarial features
- Risk scoring — 0-100 risk assessment before every execution
- Ship pipeline —
archon ship: review → test → version → changelog → PR - Visual QA —
archon qa: headless browser health checks with diff-aware page testing - Session retrospective —
archon retro: quantitative metrics on your session
Context-aware intelligence for your existing AI tools: Cursor, Claude Code, Windsurf, Amp, Copilot, Gemini, and more. Your AI detects what you're doing and proactively offers the right help — no commands to memorize.
Context-Aware Intelligence (NEW):
- Fix-First Code Review — Auto-fixes mechanical issues, only asks about ambiguous ones
- AI Slop Detection — 10-item blacklist catches generic AI design patterns, scored A-F
- Systematic Debugging — Root cause first, 3-strike hypothesis testing, regression prevention
- Ship Readiness Dashboard — Pre-deploy checklist tracks all quality gates across your session
- Scope Management — 4 modes (expand/selective/hold/reduce) when scope creep detected
- Self-Regulation Guardrails — Blast radius checks, file limits, automatic checkpoints
- Design Governance — DESIGN.md as design source of truth with visual audit
- Completion Status — Clear DONE/CONCERNS/BLOCKED/NEEDS_CONTEXT reporting
- Progress Reflection — Solo retro at natural milestones
- Expert Review Spec — Generate a consultant-ready technical spec (gap analysis, approach review, UX enhancement) at any milestone
- Features Dashboard — Live HTML page tracking which capabilities you've used, what's available, and what's recommended next
Governance Foundation:
- .archon/active/architecture.md template with best practices
- Quality Level / Posture — prototype/production/enterprise-grade behavior
- DEPENDENCIES.md — File-level dependency tracking to prevent regressions
- IDE-specific rule files (.cursorrules, CLAUDE.md, GEMINI.md, etc.)
- 24 on-demand AI skills — Design review, debugging, ship readiness, scope review, accessibility, SEO/GEO, color scheme picker, expert review, features dashboard, and more — loaded only when triggered, keeping context lean
- Skills-based architecture — 80% smaller context footprint (8.6 KB core vs 42 KB monolithic). Skills load on-demand with progressive disclosure
- Claude Code: 16 native slash commands —
/debug,/design-review,/ship-readiness,/code-review,/scope-check,/expert-review,/geo-optimize,/seo-check,/accessibility,/reflect,/handoff,/plan-tasks,/color-scheme,/rollback,/dashboard,/constitution - GEO Optimization — 7-phase protocol for AI search citation: identity phrases, atomic claims, JSON-LD schemas, and audit (free tools)
- Task Extraction — AI confirms all items before starting, nothing gets forgotten
- Context Handoff — Memory management for long sessions
- Works with any AI coding assistant
Governance data for this repo lives in .archon/governance.db and is local-only. It is never synced to Supabase.
To initialize or refresh it:
pnpm exec tsx scripts/init-governance-db.ts| Command | Description |
|---|---|
archon |
Chat-first interactive mode — just run and start talking |
archon init |
Initialize in your project |
archon mode |
Choose local governance mode or BYOK AI mode |
archon config ai |
Guided BYOK provider-key setup |
archon status |
Show local mode, auth, and project status |
archon plan <description> |
Create a work item with AI planning (extracts and confirms multi-item requests) |
archon execute <atom-id> |
Execute with quality gates |
archon list |
List all work items |
archon show <atom-id> |
Show details |
archon watch |
Live TUI dashboard with status |
archon bug report <title> |
Bug report with root cause analysis |
archon review init |
Initialize AI-powered code review |
archon review analyze |
Scan project and populate review tasks |
archon review run |
Run AI review on pending tasks |
archon usage |
Usage by period and model |
archon keys add <provider> |
Add your own API key — BYOK (Bring Your Own Key) |
archon keys list |
Show configured API keys by provider |
archon preferences |
Interactive settings menu (models, keys, usage) |
archon models |
List available AI models |
archon deps list |
View file dependency rules |
archon deps add |
Add a new dependency rule |
archon deps check --files <list> |
Check for downstream impacts |
archon deps graph |
Generate Mermaid dependency diagram |
archon a11y check |
Run WCAG 2.2 AA accessibility audit |
archon a11y fix |
Auto-fix common accessibility issues |
archon a11y badge |
Add accessibility compliance badge |
archon a11y pre-deploy |
Interactive pre-deployment check |
archon seo check |
Run SEO meta tag audit |
archon seo fix |
Apply recommended SEO fixes |
archon geo identity |
Generate brand identity phrases for AI citation |
archon geo schema |
Generate JSON-LD schemas |
archon governance status |
Show governance status (AGD) |
archon governance architecture update |
Update architecture with change reason |
archon governance task update |
Update governance tasks |
archon governance handoff |
Log handoff + current context |
archon governance migrate |
Migrate legacy governance files |
archon governance sqlite-init |
Initialize or refresh local governance SQLite DB |
archon index init |
Initialize local semantic indexing |
archon index update |
Index changed files |
archon index search "query" |
Semantic code search |
archon parallel status |
Show parallel execution status |
archon ship |
Ship pipeline: review → test → version → changelog → PR |
archon ship --dry-run |
Run all ship checks without committing |
archon qa |
Visual QA with headless browser health checks |
archon qa --url <url> |
Test specific URL |
archon retro |
Session retrospective with quantitative metrics |
archon deploy |
One-click deploy (auto-detect platform) |
archon cleanup check |
Analyze workspace for bloat |
archon cleanup run |
Execute cleanup tasks |
archon cleanup auto [enable|disable] |
Enable/disable auto cleanup on start |
Notes:
- Content-only requests (stories, outlines, lessons, visuals) use lightweight planning to avoid blocking.
- BYOK shows per-model usage and cost by today/week/month/year in
archon preferences→ “View usage details.” - You can paste multi-line requests into interactive prompts; Archon captures them as a single response.
- Proposal approvals like
approve plannow bind to the pending proposal context in chat mode. - Governance boundary/path checks in execute now steer with actionable guidance and set atoms to
BLOCKEDrather than hard failing. - Analysis-first requests now return recommendations first and accept natural confirmations (
yes,go ahead,create) to create governed tasks. - Chat continuation supports natural directives (
continue,move forward), and path-scope governance blocks now attempt automatic re-plan + retry recovery. - Chat now preserves your high-signal original request better across exploratory turns (for example:
check folders), and avoids duplicate plan/execution prompts. - After explicit approvals in chat (
yes,approve plan), execution continues directly without extra confirmation prompts.
Tip: Use archon plan --edit to adjust title and acceptance criteria before planning.
Web Checks: If Archon detects a web project, it prompts to run A11y/SEO/GEO checks and stores your preference in .archon/config.yaml.
ArchonDev is Free. There are no paid tiers, no subscriptions, and no credit card required.
| Mode | Cost to ArchonDev | What You Get |
|---|---|---|
| Download / Governance | $0 | Governance/workflow layer with no built-in AI calls |
| CLI + AI (BYOK) | $0 | Full CLI with AI — bring your own API keys; your provider bills tokens directly at cost |
No Archon token markup. You pay your LLM provider directly at their published rates.
- Your API keys are stored locally in
~/.archon/keys.json(never uploaded to ArchonDev servers). - Keys are encrypted at rest with AES-256-GCM and the file is set to owner-only permissions (
0600). - Archon uses your keys only to call your chosen providers on your behalf.
- If your device is compromised, an attacker could access local files. Treat keys as sensitive secrets.
- Run
archon— Archon inspects the current folder and governance state - Add keys only if needed — BYOK setup is local-only and optional until you want AI actions
- Start in chat — Ask naturally; Archon decides when to stay conversational vs create atoms
- Governance loads first — Architecture, dependencies, and progress are respected automatically
- Changes are validated — Quality gates and approval policy stay active
- Continuity persists — Handoffs, pending work, and learnings stay local across sessions
$ archon
ArchonDev - AI-Powered Development Governance
────────────────────────────────────────────────
AI mode: Local governance only
No provider keys configured yet.
→ Run 'archon config ai' to enable BYOK for CLI AI features
→ What kind of project are you building?
[AI asks natural follow-up questions based on your answers]
(Type "mode" or "help" anytime)
✓ Project initialized!Type these anytime during interactive prompts:
| Command | Description |
|---|---|
config ai |
Open BYOK key setup |
status |
Show local mode and optional auth info |
keys |
List configured API keys |
help |
Show available commands |
quit |
Exit ArchonDev |
These commands are preserved only for backward compatibility and are hidden from the normal CLI surface. They are not part of the current local-first product path.
# 1. Authenticate
archon login
# 2. Connect GitHub (one-time setup)
archon github connect # Opens browser for authorization
archon github status # Verify connection
# 3. Plan locally, execute in cloud
archon plan "add user settings page"
archon execute ATOM-001 --cloud
# 3b. Queue multiple atoms in parallel (legacy command; disabled in free/BYOK mode)
archon parallel cloud ATOM-001 ATOM-002
# 4. Check progress
archon cloud status # List all cloud executions
archon cloud logs <id> # View execution logsCloud execution is currently disabled in the free/BYOK model. Use local execution with BYOK keys.
Have a project created by another AI agent? ArchonDev can review it first, then govern future changes.
# Step 1: Review existing code
cd your-existing-project
archon review init # Create local review database
archon review analyze # Scan project structure
archon review run --all # AI reviews all features
archon review export > review-report.md
# Step 2: Set up governance
archon init --analyze # Creates ARCHITECTURE.md
# Step 3: Fix issues with governed workflow
archon plan "fix critical issues from review"
archon execute <atom-id>The CLI detects existing projects and suggests this workflow automatically.
- archondev.io — Main website
- archondev.io/geo — Free GEO optimization tools (Claude skill + prompt generator)
- archondev.io/color-schemes — Color Scheme Picker (26 curated schemes)
- AI Coding Problems Research — Market research on AI coding assistant issues