waii

WAII MCP Server provides a natural language interface to databases via WAII. The server exposes a single tool named 'database' that lets language models send natural language queries and have them translated into SQL or database operations under the hood. To use it, configure the server with your WAII URL, API key, and your database connection string. The model can then ask questions about schema, run queries, and receive structured results. The language-to-SQL capabilities include understanding schema structure, composing appropriate SELECT/INSERT/UPDATE/DELETE statements, and returning results in a consumable format. You can also leverage the built-in data interpretation features to understand column types and relationships, making it easier to formulate accurate queries from plain language prompts.

Once configured, you interact with the server by starting it with your preferred runner (uv/uvx, as shown in the installation instructions). The 'database' tool will accept three arguments: database-key (your connection string), api-key (WAII API key), and url (WAII API endpoint). The MCP framework will route natural language prompts to WAII, which executes on your database and returns results. This setup enables workflows such as asking for data insights, generating SQL templates, checking schema details, and performing data retrieval without writing SQL by hand.

Prerequisites:

• Python 3.10 or higher
• Access to a WAII account with API access
• Internet connection for package installation

Installation steps (recommended uv/uvx workflow):

1. Install the server package (via pip):

pip install mcp-server-waii

2. Run the server using uv/uvx (examples):

• Using uvx (recommended):

uvx run -m mcp_server_waii \
  --url "YOUR_WAII_URL" \
  --api-key "YOUR_API_KEY" \
  --database-key "YOUR_DATABASE_CONNECTION_STRING"

• Using uv (alternative):

uv run -m mcp_server_waii \
  --url "YOUR_WAII_URL" \
  --api-key "YOUR_API_KEY" \
  --database-key "YOUR_DATABASE_CONNECTION_STRING"

3. Alternatively, run as a Python module (if preferred):

python -m mcp_server_waii \
  --url "YOUR_WAII_URL" \
  --api-key "YOUR_API_KEY" \
  --database-key "YOUR_DATABASE_CONNECTION_STRING"

4. Development setup (optional):

• Clone the repository and install in editable mode:

git clone https://github.com/waii-ai/mcp-server-waii.git
cd mcp-server-waii
pip install -e .

• Run in development mode with uv/uvx as above.

Prerequisites recap:

• Python 3.10+
• Pip (package installer)
• Access to WAII (URL and API key)
• A valid database connection string compatible with WAII

Tips and notes:

• Ensure your WAII URL is accessible from the environment where the MCP server runs, and that your API key has the required permissions.
• The database-key should contain your database connection string compatible with WAII's requirements.
• If you see authentication or connectivity errors, double-check the API key and URL, and verify network access to WAII endpoints.
• The server focuses on natural language to SQL translation and may benefit from clear, well-structured prompts that reference the database schema when possible.
• The configuration example uses uvx with the mcp_server_waii module name; if you deploy differently, adjust the command and path accordingly.
• For troubleshooting, consult the GitHub Issues page or WAII support channels referenced in the repository.