Quick start
This quick start takes you from a new InfoConnect Hub to a working API, and then to the same API as a tool for an AI agent. It has four steps: install the Hub, create your first API, test it, and open it to an AI agent.
The Hub quick start covers the same first steps and adds a Program Call, data replication to a database, and a green-screen script.
Step 1. Install and start the Hub
- Check the prerequisites: IBM i V5R4 or later, network access from the Hub to IBM i, and Java 17.0.1 or later for the Hub.
- Install and start the Hub from the distribution archive. See Installing and starting the Hub and Installation, services and updates.
- Open the web interface. In the bundled all-in-one configuration it is at
http://localhost:8085.
If you prefer containers, see Deploying InfoConnect Hub with Docker, Kubernetes and Cloud Run.
Step 2. Create a connection and your first API
Follow Getting started with the web UI. It walks you through these tasks:
- Sign in.
- Create a Native Connection to your IBM i system.
- Create a Data API that surfaces an SQL query as a REST API. You need no IBM i program for this.
To surface an existing IBM i program, create a Program Call instead. See Program Call.
Step 3. Test the API
Open the API in the web interface and run it. The result shows the HTTP status, the response time, and the JSON response. See Executing APIs.
You can also call the API from any REST client. The Postman collection contains ready-made requests. To generate an OpenAPI specification for your APIs, see Managing Hub APIs.
Step 4. Open the API to an AI agent
The Hub includes an MCP server. It generates a tool for each Program Call, Data API, and RPA Call that you create.
- Make sure that the Hub and the MCP server are running.
- Open the MCP settings of your AI agent (for example Cursor or Claude) and add a custom MCP server of type
streamable-httpwith the URLhttp://localhost:8083/api/mcp. - Wait until the MCP server loads your APIs, then ask the agent a question that needs your IBM i data.
See the AI and MCP tutorial for the full configuration, the example system prompt, and the MCP Explorer, which lets you browse and test the generated tools in the web interface.
Automate a green-screen application
If you can reach the application only through its green screens, use the RPA Hub. See the RPA Hub quick start.
Next steps
- Hardening and scaling before you go to production.
- Tutorials and examples for end-to-end scenarios.