# Architect Ai > AI-powered study tool for Claude Certified Architect exam. The codebase IS the curriculum. - **Type:** MCP server - **Install:** `agentstack add mcp-aviraldua93-architect-ai` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [aviraldua93](https://agentstack.voostack.com/s/aviraldua93) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [aviraldua93](https://github.com/aviraldua93) - **Source:** https://github.com/aviraldua93/architect-ai - **Website:** https://github.com/aviraldua93/architect-ai#readme ## Install ```sh agentstack add mcp-aviraldua93-architect-ai ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # ArchitectAI — Free Study Tool for the Claude Certified Architect Exam **The codebase IS the curriculum — every file demonstrates an exam concept in production code.** [](https://github.com/aviraldua93/architect-ai/stargazers) [](LICENSE) [](https://github.com/aviraldua93/architect-ai/actions/workflows/ci.yml) [](tsconfig.json) --- ## What Is This? ArchitectAI is an open-source study tool for the [Claude Certified Architect (Foundations)](https://www.anthropic.com/certification) exam — but with a twist: **the codebase itself teaches you the exam concepts**. Every architectural decision, every file, every pattern maps directly to one of the 30 exam task statements across 5 domains. You don't just *read* about agentic loops — you see a production implementation in `src/agents/loop.ts`. You don't just *memorise* hook patterns — you see `PostToolUse` normalisation running live in `src/agents/hooks.ts`. > 🆓 **No API key required for Tier 1.** 162 practice questions, CLI quiz, timed exam simulation, spaced repetition — completely free and offline. ## Feature Status | Feature | Status | Notes | |---------|--------|-------| | CLI Quiz (`quiz`) | ✅ Shipped | 132 questions across all 5 domains | | Timed Exam Simulation (`exam`) | ✅ Shipped | 60-question mock exam, 120-min timer, domain-weighted, 720/1000 pass threshold | | Adaptive Study (`study`) | ✅ Shipped | SM-2 spaced repetition, domain filtering, source code references | | Exam Readiness (`assess`) | ✅ Shipped | Per-domain scoring, weak area identification, study recommendations | | Progress Dashboard (`dashboard`) | ✅ Shipped | Terminal dashboard with domain bars, streak tracking, SRS due count | | Web UI | ✅ Shipped | Next.js app at `localhost:3939` — quiz with filters, dashboard, dark mode, mobile-responsive | | MCP Server | ✅ Shipped | Full JSON-RPC 2.0 server with resources, tools, and prompts | | Exam Scenarios | ✅ Shipped | 6 production scenarios with 30 linked questions matching official CCA-F format | | Concept Deep-Dives | 🚧 Planned | In-depth explanations for all 30 task statements | ## Quick Start ```bash git clone https://github.com/aviraldua93/architect-ai.git && cd architect-ai npm install npx architect-ai quiz ``` Or install globally: ```bash npm install -g architect-ai architect-ai quiz ``` > 💡 **That's it.** No API key, no account, no cost. Just `quiz` and start practicing. ### More Commands ```bash architect-ai exam # Timed 60-question mock exam architect-ai study --domain 1 # Adaptive study with spaced repetition architect-ai assess # Exam readiness prediction architect-ai dashboard # Progress tracking dashboard ``` ### Web UI ```bash cd web && npm install && npm run dev # Opens at http://localhost:3939 ``` ## Three Ways to Learn ### 🖥️ Tier 1 — Static CLI **No API Key Required** 105 scenario-based questions across all 5 exam domains. Timed exam simulation. Spaced repetition. Progress tracking. Works offline. Works on a plane. **ZERO COST.** ### 🔌 Tier 2 — MCP Server **No API Key Required** Plugs into your existing Claude Desktop, GitHub Copilot, or any MCP client. Our server provides content + tools. Your AI provides intelligence. **ZERO COST.** ### 🤖 Tier 3 — Live Agents **Bring Your Own Key** Full multi-agent orchestration with adaptive study sessions, Socratic dialogue, and weakness detection. You provide your own Claude API key. **~$0.02–0.05 per session. FULL POWER.** ## Exam Coverage Every file maps to a specific exam domain and task statement: | Domain | Weight | Source Directory | Key Concepts | |--------|--------|-----------------|--------------| | **1. Agentic Architecture** | 27% | `src/agents/` | Agentic loops, multi-agent orchestration, subagent context passing, workflow enforcement, SDK hooks, task decomposition, session state | | **2. Tool Design & MCP** | 18% | `src/tools/`, `src/mcp/` | Tool descriptions, error schemas, tool_choice modes, MCP resources/tools/prompts, built-in tools | | **3. Claude Code Config** | 20% | `src/cli/`, `.claude/` | CLAUDE.md hierarchy, slash commands, path rules, plan mode, iterative refinement, CI/CD | | **4. Prompt Engineering** | 20% | `src/prompts/` | Explicit criteria, few-shot examples, structured output, validation-retry, batch API, multi-instance review | | **5. Context & Reliability** | 15% | `src/context/` | Context preservation, escalation patterns, error propagation, codebase exploration, human review, provenance | ## Usage ```bash # Tier 1: Offline study (no API key needed) npx architect-ai quiz # Quick practice questions npx architect-ai quiz -d 1 # Filter by domain (1-5) npx architect-ai quiz -t 1.2 # Filter by task statement # Tier 2: MCP server (plug into Claude Desktop / Copilot) npm run mcp # Start MCP server # Tier 3: Live agents (requires ANTHROPIC_API_KEY — coming soon) export ANTHROPIC_API_KEY=your-key # macOS/Linux # $env:ANTHROPIC_API_KEY="your-key" # Windows PowerShell npm run agent # Adaptive study session ``` ## Project Structure ``` architect-ai/ ├── CLAUDE.md ← Domain 3: Claude Code configuration ├── .claude/commands/ ← Domain 3: Slash commands (/study, /quiz, /explain, /assess) ├── .claude/rules/ ← Domain 3: Path-specific rules ├── src/ │ ├── agents/ ← Domain 1: Agentic Architecture (27%) │ │ ├── coordinator.ts Hub-and-spoke orchestration (1.2) │ │ ├── loop.ts Agentic loop lifecycle (1.1) │ │ ├── spawner.ts Subagent invocation (1.3) │ │ ├── workflow-enforcer.ts Programmatic enforcement (1.4) │ │ ├── hooks.ts PostToolUse & interception hooks (1.5) │ │ ├── decomposer.ts Task decomposition strategies (1.6) │ │ └── session-state.ts Session management & resumption (1.7) │ ├── tools/ ← Domain 2: Tool Design (18%) │ │ ├── definitions.ts Tool schemas & descriptions (2.1) │ │ └── error-handling.ts Structured error responses (2.2) │ ├── mcp/ ← Domain 2: MCP Integration │ │ └── server.ts MCP server: resources, tools, prompts (2.4) │ ├── prompts/ ← Domain 4: Prompt Engineering (20%) │ │ ├── system-prompts.ts Agent persona & constraints (4.1) │ │ ├── few-shot.ts Example banks & dynamic selection (4.2) │ │ └── output-schemas.ts Zod validation & retry loops (4.3) │ ├── context/ ← Domain 5: Context Management (15%) │ │ ├── session-manager.ts Session save/load/resume/fork (1.7, 5.1) │ │ ├── progress-tracker.ts Weighted scoring & spaced repetition (5.1) │ │ └── escalation.ts Tier 1→2→3 escalation patterns (5.2) │ ├── cli/ ← Domain 3: CLI & Commands (20%) │ └── content/ ← Pre-generated study material │ └── questions/ 105 scenario-based questions ├── web/ ← Next.js web UI (in progress) ├── test/ │ ├── tier1/ ← Free static validation (<5s) │ ├── tier2/ ← E2E with real Claude (~$3-4/run) │ └── tier3/ ← LLM-as-judge quality checks (~$0.15/run) └── docs/ ← Architecture & contributor guides ``` ## Question Banks | Domain | Questions | Status | |--------|-----------|--------| | 1 — Agentic Architecture | 30 | ✅ Complete | | 2 — Tool Design & MCP | 20 | ✅ Complete | | 3 — Claude Code Config | 20 | ✅ Complete | | 4 — Prompt Engineering | 20 | ✅ Complete | | 5 — Context & Reliability | 15 | ✅ Complete | | **Total** | **105** | | ## The Meta-Pattern This project practices what it preaches. The way it's built IS a demonstration of exam concepts: - **This README** → You're reading structured context passing (Domain 5) - **The org that built this** → Hub-and-spoke multi-agent orchestration (Domain 1.2) - **Branch protection + CI** → Programmatic enforcement over prompt-based guidance (Domain 1.4) - **3-tier testing** → Task decomposition: per-file analysis + cross-file integration (Domain 1.6) - **MCP server architecture** → Tool design with proper descriptions and error schemas (Domain 2.1, 2.2) ## Contributing Contributions welcome! See **[CONTRIBUTING.md](CONTRIBUTING.md)** for the full guide — branch naming, PR process, testing, code style, and how to add questions. ## About **ArchitectAI** is an [aviraldua93](https://github.com/aviraldua93) project — built by [Aviral Dua](https://github.com/aviraldua93) using the very agentic patterns it teaches. --- *Preparing for the Claude Certified Architect exam? Star this repo and start with Tier 1. No API key. No cost. Just learning.* ## Source & license This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [aviraldua93](https://github.com/aviraldua93) - **Source:** [aviraldua93/architect-ai](https://github.com/aviraldua93/architect-ai) - **License:** MIT - **Homepage:** https://github.com/aviraldua93/architect-ai#readme Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** no - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/mcp-aviraldua93-architect-ai - Seller: https://agentstack.voostack.com/s/aviraldua93 - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.