Model Monster MCP Walkthroughs

This repository contains walkthroughs showing you how to use the Model Monster MCP server with real AI system codebases.

The workflow is:

1. connect an MCP-capable agent to Model Monster
2. clone an example system locally
3. choose the Model Monster org and team for the system
4. ask the agent to inspect the codebase and prepare a blueprint for review
5. review and adjust the blueprint
6. tell the agent to upload the reviewed system

Use this README to understand the walkthrough repo and choose an example. Use MCP setup to configure Model Monster access.

What This Repo Is

This repo is a tutorial workspace for the Model Monster MCP/API workflow. Use it to:

• learn how agent-driven blueprint creation works
• walk through review-first blueprint creation
• compare how different AI system architectures map into Model Monster

The repo does not include a separate scanner. The agent reads the target repo and uses Model Monster MCP tools to prepare and upload the system blueprint.

Repo Layout

• examples/ — one folder per example system
• shared/ — MCP setup instructions and shared examples
• templates/ — copy-paste prompt templates for common steps
• workspaces/ — local target repositories for scanning; ignored by git

Quick start

1. Complete MCP setup.
2. Choose an example walkthrough.
3. Clone the target repository into workspaces/.
4. Choose the destination org and team in Model Monster.
5. Ask the agent to prepare a blueprint for review.
6. Review and adjust the blueprint.
7. Tell the agent to upload the reviewed system.

Example walkthroughs:

• LibreChat walkthrough
• OPEA ChatQnA walkthrough
• SRE Agent walkthrough

Current Examples

• examples/librechat/ — self-hosted AI chat platform with agents, MCP, RAG, search, and external model providers
• examples/opea-chatqna/ — OPEA microservice-based RAG application with a UI, gateway, embedding, retrieval, reranking, LLM, data prep, and vector database
• examples/sre-agent/ — AI SRE agent that reads operational logs, inspects source code, diagnoses issues, suggests fixes, and reports to Slack

Expected Output

A successful run should produce:

• a reviewed system blueprint
• a destination system in Model Monster
• a summary of key nodes, resources, edges, assumptions, and warnings

Local Files

These files are for local use and should not contain committed secrets:

• .env
• workspaces/

The checked-in MCP config examples live at:

• .mcp.json
• .vscode/mcp.json
• shared/examples/model-monster.mcp.json.example
• shared/examples/codex.config.toml.example
• shared/examples/cursor.mcp.json.example