Usage Guide

This guide provides detailed instructions on how to run and interact with Curaitor Agent after installation. It covers launching the web interface, using the MCP Inspector, managing scientific data workflows, and troubleshooting common issues.

Running the Web Interface

The Curaitor Agent includes a lightweight web application that allows you to manage tasks, upload datasets, and interact with tools through a browser-based interface.

Steps

Open a terminal and navigate to your project directory:
```
cd curaitor-agent
```
Note

Ensure you are in the root directory of your Curaitor Agent project where uv commands are available.
Start the web interface:
```
uv run adk web
```
- This command launches the web server locally.
- By default, the server runs on port 8000.
Open the web interface in your browser:
```
http://localhost:8000
```
- You should see the Curaitor Agent dashboard.
- Here, you can upload datasets, manage tasks, and use available tools.
Using a custom port:
```
uv run adk web --port 8080
```

—

Using the MCP Inspector

The MCP Inspector is a tool for testing and verifying connections between Curaitor Agent and its integrated tools.

Steps

Ensure the Curaitor Agent server is running.

Launch the MCP Inspector:

npx @modelcontextprotocol/inspector uv run tools/mcp_server.py

Connect through the MCP Inspector UI:
- Click Connect to establish a connection with the running server.
- Select Test Tools to check available commands and ensure they respond correctly.

Important

Use Node.js v22.x for compatibility. Running MCP Inspector with an unsupported Node version may lead to connection failures.

—

Example Workflow

Here is a step-by-step example workflow to help you get started:

Start the Curaitor Agent:
```
uv run adk web
```

Verify server tools with MCP Inspector:

npx @modelcontextprotocol/inspector uv run tools/mcp_server.py

Upload and process data:
- Use the web interface to upload your dataset.
- Apply tools such as parsing, filtering, or data transformations.
- Review results and export them for downstream analysis.

—

Best Practices

Keep dependencies up-to-date:
```
uv sync
```
Use version control (e.g., Git) to track changes in project setups and workflows.
Test new tools in MCP Inspector before integrating them into production workflows.
Ensure reproducibility: Run experiments in a clean environment to avoid inconsistencies.

—

Troubleshooting

Web interface fails to start: - Ensure all dependencies are installed. - Run uv sync to synchronize packages.
MCP Inspector cannot connect: - Verify that the Curaitor Agent server is running. - Check firewall settings or network rules that may block connections.
Package issues: If a tool or package fails, remove and re-add it:
```
uv remove package-name
uv add package-name
```

—

Next Steps

Once you are comfortable using Curaitor Agent:

Explore the API reference for programmatic integration.
See the MCP Inspector documentation for advanced testing.
Extend workflows for your research by combining multiple tools and datasets.
Contribute to the project by following the guidelines in the Contributing page.
Stay updated with the latest features and improvements by checking the repository regularly.
For licensing details, refer to the MIT License page.
Always ensure your environment is synchronized with requirements.txt for consistent results.