Hand work to the mabl agent in a conversation and come back to the result when it's done. Instead of clicking around in the UI, you can describe what you need in plain language, whether it's one simple change or a sequence of work, such as fixing a failing test, adding it to a plan, and running the plan. This article covers how to start a conversation, redirect the agent while it works, and pick a conversation back up later.
Start a conversation
Start a conversation from any of these places:
- The prompt on your workspace dashboard
- The Chat with mabl icon in the top bar, which opens a full-page conversation with your chat history beside it
- Build tests > Start chatting...
- A test run, plan run, or deployment event page, where the conversation starts from that result
Describe what you need. A single request can span several kinds of work, for example:
Create a QA environment for the checkout app, then add the checkout tests to the Smoke tests plan and run the plan.
To give the agent more context, click + below the prompt to attach files, existing tests, or test cases. You can send attachments without typing a message, and the agent responds to what you attached. A new conversation also suggests a few places to start, such as planning a test or investigating a failure. Click a suggestion to send it as your first message.
For tips on describing a new test, see How to write an effective browser test prompt.
Let the agent work
Ask for whatever you'd otherwise do yourself across the workspace, for example:
- Build a new test from a description, a test case, or a Jira issue
- Change an existing test, or update every test that uses a URL
- Find out why a run failed and fix the test
- Add tests to a plan, create a plan, or run one
- Set up an application or environment for a new app
- Create a DataTable for data-driven tests
- Merge a branch once its tests are ready
The agent acts with your permissions, so it can only make the changes your role allows.
You don't need to stay on the page while the agent works. The conversation keeps going after you close the tab. When the agent starts a test run or plan run, it picks the conversation back up once the run finishes, reads the results, and tells you what happened. When it starts authoring agents, it keeps track of their progress and answers their questions where it can.
Change course while the agent works
You don't have to wait for the agent to finish before you redirect it. Send a message at any point, such as "Run the Regression plan instead," and the agent reads it at its next step and adjusts. Your message shows as Queued until the agent picks it up.
Answer the agent's questions
When the agent needs something from you, it asks in the conversation, and you reply in the chat box. Authoring tasks the agent starts can ask questions too. A task's question appears at the top of its Overview tab, where you can answer it or reply in the chat box below. If the agent asks something new later, the question opens again so you don't miss it.
Only the person who started an authoring task can accept, close, or stop it. Teammates can still open the task and follow along. For task statuses and notifications, see Agent tasks and notifications.
Come back to a conversation
Your conversations are private. Other members of your workspace can't see them, and you can't see theirs. To find one again:
- On your workspace dashboard, look under Pick up where you left off, or select View all.
- In the full-page chat, use the Chat history beside the conversation. Search it, or select Start new to begin another conversation.
Hover over a conversation in the chat history to pin, rename, or archive it. Archiving removes the conversation from your history and stops any work the agent is still doing in it, and you can't continue an archived conversation.
When a conversation gets too long to continue, the agent says so and asks you to start a new one. Your tests and runs aren't affected.