AI Agents for Developers – Demo Project

This repository contains three example projects demonstrating the use of multi-agent AI systems and admin tooling for developers. Each example is self-contained and showcases a different use case or integration.

Examples

1. create_lint_rule

A multi-agent system (powered by crewAI) that demonstrates how agents can collaborate to create a lint rule. The system is highly configurable, allowing you to define your own agents and tasks.

• Directory: create_lint_rule
• Key Features:
  • Multi-agent collaboration
  • Easily extensible via YAML config files
  • Generates a report.md as output

2. docs_updater

Another crewAI-based multi-agent system, this one focused on updating documentation. Like the lint rule example, it is highly configurable and outputs a report.md file.

• Directory: docs_updater
• Key Features:
  • Multi-agent collaboration
  • Customizable agents and tasks
  • Designed for documentation workflows

3. app-admin-mcp

A server providing admin functionalities for a CRM app, including note storage, summarization, and note-adding tools. Integrates with the Model Context Protocol (MCP) and is suitable for use with tools like Claude Desktop.

• Directory: app-admin-mcp
• Key Features:
  • Note storage and summarization
  • MCP server integration
  • Example configuration for Claude Desktop

---

Installation

All examples require Python 3.10–3.13 and use uv for dependency management.

1. Install uv (if not already installed):

pip install uv

2. Install dependencies for each example:

cd <example-directory>
uv sync

For the create_lint_rule and docs_updater examples, you also need to install the Javascript dependencies:

cd <example-directory>
npm install

3. Set up environment variables:

• Add your OPENROUTER_API_KEY to the .env file in each example directory (if required).

---

Running the Examples

docs_updater

The doc updater example is a simple crewAI-based multi-agent system that updates a documentation website.

You can start the demo application by running:

npm run dev

This will start the dev server on http://localhost:5173.

You can run the docs (which are built with VitePress) by running:

npm run docs:dev

This will start the docs server on http://localhost:5174.

You can then edit the app (in src/App.tsx) and start the doc updater by running:

uv run src/docs_updater/main.py

create_lint_rule

The create lint rule example is a crewAI-based workflow that creates an eslint rule.

You can run it by running:

uv run src/create_lint_rule/main.py "<description of the rule>"

You can also start the MCP server by running:

uv run src/create_lint_rule/mcp_server.py

Then, you can then add this server to your Cursor editor:

{
  "mcpServers": {
    "create-lint-rule": {
      "url": "http://localhost:8000/sse"
    }
  }
}