Qalion AI
Qalion AI answers questions about your project and, when you let it, drafts test cases, explores your site, runs tests and triages problems. Nothing changes until you approve it.
Where to find it
- Qalion AI in the sidebar opens the full page: your chats (New chat, Pinned, Recents, Search chats), the conversation, and a Memory and skills panel you open from the rail.
- On other pages, the round chat button on the right edge (Qalion Chat on hover) opens a floating window. You can detach it and move it anywhere.
- Shortcuts: Ask Qalion AI on the Dashboard, Explain this trend on Insights, Generate with AI on Test cases.
Qalion AI knows which page you are on and what is on screen. A new chat works on the project selected in the sidebar, and never acts on another project. Choose "No project — general testing advice" for questions that don't need your data.
Owners, Admins and Testers can use Qalion AI.
Ask and Agent modes
Pick the mode from the composer's mode menu (its button shows Agent or Ask):
| Mode | What it does |
|---|---|
| Ask | Answers questions about your project. It reads and answers, and never changes anything. |
| Agent | Also proposes changes, which you approve on their cards. |
Read the project (on by default) lets it read your test cases, documentation and run history. Allowed in this chat lists what you have allowed so far, each with Revoke.
Review cards
Every write, run, exploration and deletion stops on a card that previews it: the case changes, the environment, the browsers, the number of sessions. Reading needs no card.
- Allow once approves this action. Allow in this chat also lets later actions of the same kind run in this chat without asking; revoke it from Allowed in this chat in the same menu.
- For a batch: Approve or Reject each item, Reject all, then Apply decisions.
- Edit changes the proposal before you approve it (Save and approve).
- When you reject, you can give a reason. Qalion AI reads it before it tries again.
- Deletions are asked every time and can never be allowed for the whole chat.
- Qalion AI may also ask you a question on a Qalion AI asks card.
Kinds of actions
| Kind | Covers |
|---|---|
| Write new test cases | Saves test cases and plans |
| Edit existing tests | Rewrites steps, selectors and data |
| Manage requirements | Imports tickets as requirements and proposes links |
| Write project documentation | Rewrites the project documentation the team reads (use with care) |
| Explore a website | The web agent looks at your site in a real browser; uses a session and credits (use with care) |
| Run tests | Runs test cases or a plan for real; spends credits (use with care) |
| Triage problems | Changes a problem's state or owner |
| Save project memory and skills | Notes and skills Qalion AI uses for the whole team |
| Configure the project | Environments, browser profiles and the execution time limit |
| Write to connected tools | Creates and updates Jira tickets, comments and statuses (use with care) |
| Delete | Removes something. Asked every time, never allowed for the whole chat. |
What it can do
- Test cases, folders, shared steps and datasets: read, create, edit, move, assign and delete.
- Plans: create, clone, add or remove cases, configure, schedule and run.
- Runs: read results and progress, compare runs, re-run failures and cancel.
- Problems: read, change state or owner, merge, quarantine, analyse with AI and file a Jira ticket.
- Requirements: create, import from Jira, sync, suggest links to cases.
- Configuration: environments, browser profiles and the execution time limit.
- Connected tools: the tools of the MCP servers your organization connected under Integrations.
Cases in review
Cases Qalion AI writes get the status In review, and the Qalion AI can't make them Active itself. Find them in the Written by AI · review smart view, in the folder summary ("… written by the agent, awaiting review") and on the Dashboard under Awaiting review.
To accept a case, open it and set Status to Active. To turn it down, set it to Rejected. Cases in review don't run.
Website explorations
Before writing tests, Qalion AI can ask the web agent to explore your site: open a page, follow a flow, note the fields and the wording. You approve it on an Explore a website card that shows the task, the start page, the browser and the sessions available.
- An exploration is not a test. It has no verdict and never appears in runs, insights, problems or pass rates.
- It never buys, submits data or deletes anything.
- It takes from 2 to 10 minutes at most (the project's time limit, capped), uses one browser session and spends credits like a run.
- You watch each action live, then get an outcome (Completed, Partly answered or Blocked) with its observations. Qalion AI may then propose test cases on their own cards.
Project documentation
Qalion AI reads your project documentation when it drafts test cases and answers questions. Add it in Materials → Documentation: drop specs, user stories or API docs in Markdown, JSON, CSV, YAML, XML, HTML or plain text (up to 10 MB per upload).
Composer and commands
- A model picker chooses the tier that answers: Qalion Swift (fastest), Qalion Core (balanced) or Qalion Deep (the most careful reasoning).
- Thinking, a slider in the same menu (Low, Medium, High), sets how hard Qalion AI thinks for every message until you change it: higher is slower and uses more credits.
- Attach a file: up to 5 text files per message (70 KB in total), or up to 4 images of 5 MB each (PNG, JPEG, WebP, GIF) with a tier that reads images.
- Slash commands:
/new,/compact,/memory,/skills, and/followed by a skill's name. - Long chats are summarised automatically when the context fills up. Compact now does it on demand.
Credits
Every Qalion AI answer uses credits, shown under the answer and as a total per chat. When the balance runs out, Qalion AI stops before its next step and keeps what it had written.
Was this page helpful?