The aworld MCP server suite exposes a set of AI-assisted data and API tools designed to support the AWorld project. You can run a Flask-based main server for health checks and traditional REST endpoints, a FastAPI-based service for fast, async operations, and a GAIA runner for orchestration of GAIA-driven tasks. Typical endpoints include a health check at /health, a Google API + BeautifulSoup web-scraping workflow at /search, a Deep Researcher powered query at /search/agentic, OpenRouter model interactions at /openrouter/completions, and a browser automation workflow at /browser_use. Use these endpoints to perform search tasks, fetch and parse web content, and leverage LLM-assisted tooling within your application stack.

Prerequisites:

• Python 3.12 or higher
• uv package manager installed (see uv installation steps)

Installation steps:

1. Install uv (if not already installed):

   macOS/Linux

   curl -LsSf https://astral.sh/uv/install.sh | sh

   Windows

   powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

   Or via pip

   pip install uv

2. Clone the repository: git clone https://github.com/your-username/aworld-mcp-servers.git cd aworld-mcp-servers

3. Install dependencies and sync the project: uv sync

4. (Optional) Activate the virtual environment: source .venv/bin/activate # macOS/Linux

   or

   .venv\Scripts\activate # Windows

Running the application:

• Main Flask server: uv run aworld-server

• FastAPI server: uv run aworld-fastapi

• GAIA runner: uv run gaia-runner

Alternative: run as module uv run python -m src.main

Notes and tips:

• Replace placeholder values (e.g., API keys, endpoint URLs) in the request payloads with your actual credentials.
• The repository provides multiple services; ensure you start the appropriate server for your workflow (e.g., aworld-server for health checks, aworld-fastapi for async API interactions, gaia-runner for GAIA tasks).
• If you encounter environment issues, verify that Python 3.12+ is active in your shell and that uv is properly installed.
• When running in containers or CI, map deployed host/port values in your curl payloads to the correct endpoints and avoid hardcoding localhost unless appropriate.
• Review the provided curl examples in the README to understand required fields for each service (e.g., api_key, cse_id, llm_model_name, serper_api_key).
• Ensure network access to external services (Google API, Serper, OpenRouter) is permitted in your environment and that API quotas are respected.