The PBS MCP server provides a standalone interface to the Australian PBS API for model-assisted workflows. It exposes tools that let MCP-enabled clients query PBS data such as medicine listings, pricing, prescribers, and item overviews. The server supports both standard input/output (stdio) transport for tight integration with local AI agents, as well as an HTTP mode with Server-Sent Events (SSE) and REST-style tool invocations for remote clients. To use it, start the server in stdio mode for direct tool calls from an LLM or local client, or run in HTTP mode to expose endpoints like /health, /tools, /sse, and /api/:toolName for programmatic access. Configuration of tools is documented in the client integration example, which shows how to wire an MCP client to the server and specify authentication keys if required.

Prerequisites:

• Node.js and npm installed on your machine
• Git to clone the repository

Install and run:

1. Clone the repository git clone <repository-url> cd pbs-mcp-standalone

2. Install dependencies npm install

3. Build the project npm run build

4. Start the server (stdio mode by default) npm start

Optional HTTP mode: 5) Start the HTTP server (SSE enabled) npm run start:http

or with a specific port

./start.sh http 3000

CLI usage: 6) Use the built-in CLI to query PBS data npm run cli -- <command>

Example

./start.sh cli info

Notes and tips:

• This server is designed to work with the PBS public API; if you have private API access, you can adapt the endpoints accordingly.
• Be mindful of PBS rate limits. Consider caching results or performing periodic updates to minimize repeated calls.
• The server supports both stdio and HTTP transport layers. Use stdio for tight, local LLM integration and HTTP for remote clients or tooling.
• When configuring clients, you may need to provide a PBS API subscription key or other auth credentials via environment variables or config files.
• The CLI and HTTP endpoints provide examples for common PBS API endpoints such as prescribers and item-overview. See the README for parameter formats and options.
• If you upgrade, reuse the existing mcp_config structure to ensure clients can discover and use tools without code changes.