myai

Your personal AI workflow system for developers — structured workflow, persistent project memory, and curated agents that work across Claude Code, OpenCode, Gemini CLI, Codex, and Cursor.

npx @thinhpham/myai init

What It Is

Most AI coding tools are powerful but stateless — every session starts from scratch, plans live in chat history, and context gets lost. myai fixes that.

It installs a structured workflow system into any project:

• Persistent memory — docs/ brain layer: STATE.md auto-injected every session; deeper docs (code-standards, system-architecture, etc.) read on-demand by task type
• Structured workflow — brainstorm → plan → validate → build → archive, not just "chat and hope"
• 14 agents — brainstormer, planner, researcher, fullstack-developer, code-reviewer, tester, debugger, ops-manager, and more
• 30 skills — domain expertise for React, Next.js, NestJS, databases, auth, payments, mobile, AI, and more
• 8 slash commands — project lifecycle from first idea to shipping
• 5 hooks — session-init loads your project state; privacy-block protects sensitive files
• Multi-runtime — one install, works with Claude Code, OpenCode, Gemini CLI, Codex, and Cursor

Quick Start

# 1. Initialize a project
cd your-project
npx @thinhpham/myai init              # installs .myai/ system, docs/ brain layer, plans/templates/

# 2. Wire your runtime (pick one or combine with init)
npx @thinhpham/myai install --claude   # Claude Code: copies agents/skills/hooks into .claude/, creates CLAUDE.md
npx @thinhpham/myai install --opencode # OpenCode: copies system into .opencode/
npx @thinhpham/myai install --gemini   # Gemini CLI: creates GEMINI.md
npx @thinhpham/myai install --codex    # Codex: creates AGENTS.md
npx @thinhpham/myai install --cursor   # Cursor: creates .cursor/rules/main.mdc
npx @thinhpham/myai install --all      # all five runtimes at once

# Shortcut: init + install in one step
npx @thinhpham/myai init --claude

# 4. Open Claude Code — session-init hook loads your project state automatically

# 5. Use slash commands in Claude Code
/myai:new-project      # guided setup: creates PROJECT.md, SPEC.md, ROADMAP.md, STATE.md
/myai:plan [task]      # create an execution plan with skill auto-detection
/myai:validate         # interview-based validation — confirm open questions before building
/myai:cook             # execute the active plan
/myai:pause            # save session state
/myai:resume           # restore session from last pause
/myai:verify           # verify phase goal achievement after build
/myai:archive          # journal + archive completed task plans
/myai:progress         # check project status + route to next action
/myai:kanban           # visual plans dashboard
/docs:init             # generate deeper docs/ from codebase analysis (first time)
/docs:update           # sync docs/ based on recent code changes

Commands

| Command | Description | |---------|-------------| | npx @thinhpham/myai init | Install .myai/ system + docs/ brain layer + plans/templates/ into current project. Does not wire any runtime — run install after. | | npx @thinhpham/myai init --claude | Init + install for Claude Code in one step | | npx @thinhpham/myai init --all | Init + install for all runtimes in one step | | npx @thinhpham/myai install --claude | Copy agents/skills/commands/hooks into .claude/, merge hooks into settings.json, create CLAUDE.md | | npx @thinhpham/myai install --opencode | Copy system into .opencode/, create opencode.json | | npx @thinhpham/myai install --gemini | Create GEMINI.md in project root (or ~/.gemini/GEMINI.md with --global) | | npx @thinhpham/myai install --codex | Create AGENTS.md in project root (or ~/.codex/AGENTS.md with --global) | | npx @thinhpham/myai install --cursor | Create .cursor/rules/main.mdc in project root (project-scoped only) | | npx @thinhpham/myai install --all | Install for all five runtimes | | npx @thinhpham/myai install --global | Install to user-level config dirs instead of project dir | | npx @thinhpham/myai update | Interactive merge of latest system files | | npx @thinhpham/myai validate [phase] | Print a plan phase to validate it matches requirements (defaults to latest plan) | | npx @thinhpham/myai validate [phase] --plan <name> | Validate a phase from a specific plan folder |

Slash Commands (in Claude Code)

| Command | Description | |---------|-------------| | /myai:new-project | Initialize docs/ brain layer for a new project | | /myai:plan [task] | Create a plan for a feature, fix, phase, or parallel multi-agent execution | | /myai:cook | Execute active plan — sequential by default, --parallel spawns concurrent agents | | /myai:progress | Project-wide status + smart routing to next action | | /myai:pause | Save session state to plans/{active}/PROGRESS.md | | /myai:resume | Restore session from PROGRESS.md + STATE.md | | /myai:validate [plan] | Pre-build: interview-based validation — confirms open questions and decisions | | /myai:verify [phase] | Post-build: goal-backward verification — checks phase goal achievement | | /myai:archive [plan] | Write journal entries and archive completed task plans | | /myai:kanban | Visual plans dashboard | | /docs:init | Generate all deeper docs/ files from codebase analysis (first time) | | /docs:update [scope] | Sync docs/ files based on recent code changes |

Project Structure

After npx @thinhpham/myai init:

your-project/
├── .myai/              # System files (managed by myai)
│   ├── agents/         # 13 specialized agents
│   ├── commands/       # 9 slash commands
│   ├── hooks/          # 5 session hooks
│   ├── rules/          # 4 rule files
│   ├── skills/         # 30 domain skills
│   └── settings.json   # Hook configuration
│
├── docs/               # Project brain (yours — myai never overwrites)
│   ├── PROJECT.md           # Vision, goals, tech stack, constraints
│   ├── SPEC.md              # v1/v2/out-of-scope requirements
│   ├── ROADMAP.md           # Phases with success criteria
│   ├── STATE.md             # Active plan + decisions + blockers (auto-injected)
│   ├── codebase-summary.md  # High-level codebase overview (on-demand)
│   ├── code-standards.md    # Coding conventions and patterns (on-demand)
│   ├── system-architecture.md  # Architecture and service design (on-demand)
│   ├── design-guidelines.md    # UI/UX patterns and styling rules (on-demand)
│   └── deployment-guide.md     # Deploy, CI/CD, infra instructions (on-demand)
│
└── plans/              # Execution plans (yours)
    └── {date}-{slug}/
        ├── plan.md     # Task list
        ├── PROGRESS.md # Session state (written by /myai:pause)
        └── SUMMARY.md  # Completion record

Skills

Skills activate automatically based on task content:

| Category | Skills | |----------|--------| | Core | planning, research, sequential-thinking, context-engineering, debug, fix, scout, code-review | | Frontend | frontend-development, react-best-practices, web-frameworks, ui-styling, frontend-design | | Backend | backend-development, databases, better-auth, devops | | Testing | web-testing | | Mobile | mobile-development | | AI | ai-multimodal, mcp-builder | | Payments | payment-integration | | Utility | docs-seeker, git, repomix, plans-kanban, mermaidjs-v11, brainstorm |

Multi-Session Support

State is split across two levels so multiple concurrent sessions never conflict:

• docs/STATE.md — project-level: phase focus, key decisions, blockers (≤80 lines, injected every session)
• plans/{plan}/PROGRESS.md — session-level: current task, done/remaining, next action

Each feature branch gets its own plan. Pause one, start another, resume either — no conflicts.

Supported Runtimes

| Runtime | Installation | Features | |---------|-------------|----------| | Claude Code | myai install --claude | Full: hooks, agents, skills, slash commands | | OpenCode | myai install --opencode | Agents + instructions | | Gemini CLI | myai install --gemini | Instructions (GEMINI.md) | | Codex | myai install --codex | Instructions (AGENTS.md) | | Cursor | myai install --cursor | Instructions (.cursor/rules/main.mdc) |

Requirements

• Node.js ≥ 18
• Claude Code (for full feature set)

License

MIT