Build a custom source app with an AI coding agent
Custom source apps are designed to be built with AI coding agents. Describe the data you want to analyze to an AI coding agent, such as Claude Code, Codex, or Rovo Dev. The agent builds, tests, and installs the app with the eazyBI custom source apps skill and the eazyBI CLI.
How the agent builds an app
The eazyBI custom source apps skill describes the files of an app, the JavaScript API, the data model, the tests, and the rules it must follow. The agent reads the skill and the eazyBI CLI documentation, so you can describe what you want to analyze in plain language. The skill also works with any agent that supports agent skills, and developers can read it as well.
The eazybi_custom_sources GitHub repository contains reference implementations of several custom source apps, for example, for GitHub, Help Scout, and the Atlassian Marketplace. You can install these apps as they are, or ask the agent to use them as examples.
In an account, you can also click Build new with AI on the Custom Source Apps tab and copy the instructions for the coding agent from there.
Prerequisites
- Install the eazyBI CLI version 0.5.0 or later. Check the installed version with
eazybi --version. Log in to your eazyBI site with the write scope. If you are a site admin and want to install apps for the whole site, request the admin scopes as well:
# Account owner, installs into own accounts eazybi auth login --scopes read,write # eazyBI for Jira Cloud site eazybi auth login -s example.atlassian.net --scopes read,write # Site admin, installs for the whole site eazybi auth login --scopes read,write,admin:read,admin:write
Install the eazyBI CLI skill for your agent, so that the agent knows how to use the
eazybicommand:eazybi skill install # Claude Code eazybi skill install --agent codex # Codex eazybi skill install --agent rovodev # Rovo Dev
- Create a new, empty eazyBI account for testing. You need to be the owner of the account to install a custom source app into it.
Set up a project
Create a new directory for the app and initialize a Git repository in it. Then add an AGENTS.md file to the root of the directory. It tells the agent to read the custom source apps skill before it changes anything:
# Custom source apps Read the eazyBI custom source apps skill at https://eazybi.com/skills/eazybi-custom-sources and follow it in this repository.
If you want to install the skill locally, download it as a ZIP file from https://eazybi.com/skills/eazybi-custom-sources.zip and extract it into the skills directory of your agent. The skill is updated frequently. If you install it locally, download it again regularly to get the latest version.
Sample prompt
Start the agent in the project directory and describe the data you want to import and the questions you want to answer. Below is a sample prompt for an app that imports GitHub Actions workflow runs:
Build an eazyBI custom source app that imports GitHub Actions workflow runs and jobs from the GitHub REST API for the repositories that I select in the import options. Follow the eazyBI custom source apps skill and use the github app from https://github.com/eazybi/eazybi_custom_sources as an example. I want to analyze the number of workflow runs, their duration, and their failure rate by repository, workflow, branch, conclusion, and time. First, propose the cube dimensions and measures and wait for my confirmation. Then implement the app, write tests for it, and run them with the eazyBI CLI. When the tests pass, install the app into my eazyBI test account 1234.
Tips for better results:
- Describe what you want to analyze, not only which data to import. The agent designs the dimensions and measures based on your questions.
- Ask the agent to propose the data model first. It is easier to change the dimensions and measures before the importer is written.
- Provide a link to the API documentation of the source system if it is not well-known.
- Review the agent's changes and commit them to the Git repository after each working step.
Test the app
The agent uses the following eazyBI CLI commands to test the app. You can run them yourself as well:
# Run the tests of the app on eazyBI against mocked responses, nothing is installed eazybi custom-source-apps run-tests ./github_actions # Install the app into the test account 1234 eazybi custom-source-apps install ./github_actions -a 1234 # Repeat the import of source application 56 after a change and read its log eazybi source-applications start-import 56 -a 1234 eazybi source-applications last-import-log 56 -a 1234
After the first install, add the app as a new source application in the Source Data tab of the test account, enter the connection details, and start the first import. After that, the agent can repeat the imports and read the import log with the CLI to find and fix problems.
When the app works as expected, enable it for all users. See Enable the app.
Always test a custom source app in a new, empty account. A change to the data model may require you to delete and reimport the cube.
See also
- eazyBI custom source apps skill – the complete reference of the files, the JavaScript API, and the tests of a custom source app.
- eazybi_custom_sources – reference implementations of custom source apps.
- eazyBI CLI – install the CLI and log in to eazyBI.
- Install and manage custom source apps – install, enable, update, and disable custom source apps.