Plugins
A plugin captures what your developers do in their own tools and sends it to QI Home, where Arcus generates test cases from it.
There are 2 kinds. A coding tool plugin installs in Claude Code, Cursor, Codex, or GitHub Copilot, and captures prompts, tool calls, and file changes as a session happens. The Arcus GitHub App connects your repositories and captures every pull request your team raises.
Prerequisites
- An active project in Arcus.
- The tool you are connecting, installed.
Inside Arcus, every connection is visible in one place: on QI Home, click the more options icon (⋮), then Manage integrations. Connected plugins show Active under Dev Tools and AI, and each tool’s setup link opens its install instructions.
Connect a coding tool
Section titled “Connect a coding tool”Each coding tool follows the same 3 steps: install the plugin, sign in, and set the active project. The plugin sends nothing until you sign in.
| Tool | Setup |
|---|---|
| Claude Code | Claude |
| GitHub Copilot CLI | GitHub Copilot |
| Codex CLI | Codex |
| Cursor or VS Code | Cursor |
Context and test cases go to the active project, so set it before you push.
Connect your repositories
Section titled “Connect your repositories”The Arcus GitHub App connects your GitHub repositories and sends every pull request your team raises to Unlinked Contexts. A pull request whose title or description carries a ticket ID from your connected tracker is mapped to that sprint automatically.
Use the commands
Section titled “Use the commands”The commands below are the Claude Code and GitHub Copilot forms. Cursor uses a hyphen instead of the colon (/arcus-login). Codex uses the sign-in, project, and map commands, and writes and pushes test cases when you ask it to.
| Command | What it does |
|---|---|
/arcus:help | Lists every Arcus command with a short explanation |
/arcus:login | Signs you in to Arcus so it can capture your sessions. Claude Code first asks for your region |
/arcus:logout | Signs you out and deletes the Arcus credentials stored on your machine |
/arcus:project list [search] | Lists the Arcus projects you can access |
/arcus:project use <project_id> | Sets the active project. Later events are attributed to it |
/arcus:project current | Shows the pinned project |
/arcus:map ticket <KEY> | Links a ticket (Jira, Azure DevOps, Linear, ClickUp, or GitHub) to the session, so it lands in the right sprint |
/arcus:test [feature] | Authors end-to-end test cases for the current repository via the CLI, validates them, and prompts you to run them |
/arcus:push | Pushes the test cases authored this session to Arcus, into a sprint or left unlinked |
Map what arrives
Section titled “Map what arrives”Captured work doesn’t become test cases until it is mapped:
- A session or pull request lands on the Unlinked Contexts tab of Context Management, unless it was mapped automatically. Mapping it to a sprint or an Adhoc session triggers generation from it.
- A test case pushed with
/arcus:pushlands on the Unlinked Test Cases tab. Linking it moves it into a sprint story, accepted, with its local run results carried over.
The mapping and linking flows are in Context Management.
Control what is captured
Section titled “Control what is captured”A plugin transmits nothing until /arcus:login runs. After login it sends the session’s prompts, tool calls, and captured file contents to Arcus. Sensitive credential files are always blocked (.env*, private keys, and anything under ~/.ssh, ~/.gnupg, or cloud credential stores), and dependency, build, and VCS directories are never captured. Claude Code, GitHub Copilot, and Cursor store your login token in the OS keychain. Codex stores it in a file only you can read, ~/.arcus/refresh.token, and also in the keychain if the optional keyring Python package is installed.
Each plugin’s repository lists its full capture scope and its environment-variable controls.
Was this page helpful?
Thanks for the feedback.