SSB

The SSB MCP Server provides Claude Desktop with programmatic access to SQL Stream Builder (SSB) through a comprehensive MCP API surface. It supports both direct SSB access and Knox-based authentication for CDP deployments, and is read-only by default to safely explore streams, jobs, schemas, connectors, and configuration. With 80+ MCP tools, you can manage jobs, monitor performance, inspect streams, explore tables and schemas, work with connectors and data formats, manage environments and projects, handle API keys, and perform various configuration and synchronization tasks. The server exposes endpoints via Claude Desktop so you can query SSB resources, run SQL, inspect job histories, and retrieve sample data from running streams, all within the same MCP interface.

To use it with Claude Desktop, configure the MCP server entry to point to the server (for example, via the JSON snippet in Claude Desktop’s config). If you are using a standalone SSB deployment, authenticate directly against SSB using your credentials; if you are in a CDP environment, provide a Knox token or cookie token for authentication. Once configured and the Claude Desktop application is restarted, you can start asking questions about your SSB streams, list available jobs, view histories, and execute SQL queries with real-time results where supported.

Prerequisites:

• Python 3.8+ installed on your machine or server
• Access to a running SSB instance (standalone or CDP with Knox)
• Claude Desktop (for the recommended integration with GUI tooling) or Cloudera AI Agent Studio for programmatic usage

Install steps:

1. Clone the MCP server repository: git clone https://github.com/your-org/ssb-mcp-server.git cd ssb-mcp-server

2. Create and activate a Python virtual environment: python3 -m venv .venv source .venv/bin/activate

3. Install the MCP server in editable mode (from repository): pip install -e .

4. Configure the MCP server entry in Claude Desktop or your chosen orchestration:

   • For Claude Desktop (local): set the mcpServers.ss b-mcp-server entry to: { "command": "/FULL/PATH/TO/SSB-MCP-Server/.venv/bin/python", "args": ["-m", "ssb_mcp_server.server"], "env": { "MCP_TRANSPORT": "stdio", "SSB_API_BASE": "https://ssb-host/ssb/api/v1", "SSB_READONLY": "true" } }

5. If using CDP or Cloudera AI Agent Studio, follow the environment-variable setup in the README to provide SSB base URL, Knox token, and read-only settings. You can adapt the example to your orchestration tool by setting the appropriate command and environment variables as described in the documentation.

6. Start Claude Desktop or your agent, and verify the MCP server appears in the list of available MCP servers. You should be able to query SSB resources and begin exploring via the MCP interface.

Tips and considerations:

• The MCP server is designed to be read-only by default to reduce the risk of unintended changes to SSB configurations. You can toggle the read-only flag via SSB_READONLY in the environment variables or configuration file.
• For CDP deployments, Knox-based authentication is supported. Ensure you provide a valid Knox_TOKEN or Knox cookies and the correct SSB_BASE_URL/tenant values.
• If you encounter SSL certificate issues in Knox mode, you may need to supply KNOX_VERIFY_SSL=false or provide a CA bundle path via KNOX_CA_BUNDLE.
• The configuration supports both CLI-based and GUI-based usage. In Claude Desktop, you typically provide a Python path to run the server module; in Cloudera AI Agent Studio, you can use uvx with a Git URL to run the server from source.
• Environment variables take precedence over JSON configuration. Keep sensitive values (tokens, passwords) secure and avoid hard-coding them in plaintext where possible.
• If you update SSB_BASE_URL or SSB_API_BASE, ensure they point to the correct SSB gateway or API endpoint and that network policies allow access from the MCP host.