Build Your Own Claude Code System

Explore PRDs

Paste one of these into your AI coding agent to build the entire system

Prerequisites

Docker Desktop Python 3.11+ Node.js 20+ Git Ollama

Docker runs Qdrant (vector DB) and n8n (workflow automation). Ollama provides local embeddings via nomic-embed-text. Python powers the MCP memory server, governance hooks, and dashboard. Node.js is needed for Claude Code plugins, hooks, and MCP tooling.

Claude Code

Read every PRD linked from https://platinumgoose.ai/code/ and build out all eight systems in this repository. For each PRD, implement every REQ-numbered requirement. Use the "Prompt to Build It" section in each PRD as your starting specification, but read the full Architecture Overview and Key Components sections to understand how the systems integrate. Build in this order: plugins, context-guard, memory-system, agents, governance, memory-dashboard, markdown-for-agents, code-hardener. After each system, verify it works before moving to the next.

Gemini CLI

Fetch and read every PRD linked from https://platinumgoose.ai/code/ using your web browsing capabilities. These PRDs describe eight interconnected systems. Adapt each system's architecture to work with Gemini CLI's tooling — use GEMINI.md instead of CLAUDE.md for context management, adapt hook patterns to Gemini's extension model, and replace any Claude-specific MCP configurations with equivalents that work in your environment. Implement all REQ-numbered requirements from each PRD. Build in dependency order: plugins, context-guard, memory-system, agents, governance, memory-dashboard, markdown-for-agents, code-hardener. Verify each system before proceeding.

Codex CLI

Fetch and read every PRD linked from https://platinumgoose.ai/code/ — these describe eight systems that form an autonomous development platform. Adapt each system to work within Codex CLI's architecture — use AGENTS.md instead of CLAUDE.md for context, translate hook patterns to Codex's event model, and map MCP tool integrations to Codex-compatible equivalents. The PRDs are agent-agnostic at the architecture level; the specific file formats and hook APIs need adaptation. Implement all REQ-numbered requirements. Build in order: plugins, context-guard, memory-system, agents, governance, memory-dashboard, markdown-for-agents, code-hardener. Test each before moving on.

Eight Architectural Domains

Plugin Ecosystem

8 hook events

Build extensible plugin architectures with lifecycle hooks, slash commands, and MCP integration. Transform Claude Code from a CLI into a programmable platform.

Read PRD →

Multi-Agent Orchestration

27 specialized agents

Design tiered agent workflows with 5-signal classification, quality gates, and intent sensitivity. Route tasks to the right agent with confidence.

Read PRD →

Context Management

Session persistence

Master CLAUDE.md hierarchies, auto-memory integration, and session intelligence. Maintain context across conversations and projects.

Read PRD →

Persistent Memory System

Vector-backed knowledge

Implement cross-session memory with Qdrant vectors, MCP tools, and semantic recall. Your agent remembers what matters.

Read PRD →

Agent Governance

TRiSM-compliant

Deploy trust, risk, and security management frameworks. 15 audit event types, LLM threat detection, and approval gates for production AI.

Read PRD →

Code Assurance Platform

67 integrated tools

Unified code scanning across 6 pillars. 12 scan profiles, 6-stage finding enrichment, 1000-point quality scoring, and cryptographic attestation.

Read PRD →

Why PRDs, Not Downloads?

We started by sharing agent files and plugin bundles. But we learned something critical: every environment is different. Your server topology, your security policies, your MCP tools, your workflow patterns—they're unique to you.

Instead of giving you our implementation, we're teaching you the architecture. Each PRD breaks down requirements, design patterns, and technical specifications. Then you use Claude Code to generate an implementation tailored to your environment.

You're not downloading someone else's system. You're building your own.