The Space Agent is the AI assistant built into Trace.Space. Ask it to find items, explain terms in the context of your own data, summarize a document, write or rewrite requirements, link items, or build a filter or analysis. Whenever it wants to change something in your organization, it shows you a proposal first and waits for your approval.
Opening the chat
Click the AI button in the bottom-right corner of any page. The chat opens in the right sidebar under the title Space Agent. The header has a clock icon that opens the chat history, a maximize icon that makes the chat full width, and a close icon.
On the dashboard, the My feed tab also has a chat panel headed What do you want to do today?. Type there and press Enter to start a new chat, which opens in the sidebar.
If you do not see the AI button, your user group does not have AI chat permission, or your organization has no default AI model set. An org admin sets the default model under Organization settings > AI Providers; see AI providers.
Starting a chat
A new chat shows a few example prompts under Examples:. Click one to place it in the input box, then edit it and press Enter to send. Shift+Enter inserts a new line. You can also type / in the input box to pick from the same suggestions at any time.
Giving the agent context
The item you have open is always in context. To point the agent at specific items:
Type
@in the input box and pick an item. It appears in your message and is added as a context tag above the input.Click Add context at the top of the input box to pick an item from the tree. Click the x on a tag to remove it.
Choosing a model and attaching images
The footer of the input box has a model selector. It lists the models your org admin has enabled in AI providers, grouped by provider, with the organization's default model preselected. Picking a different model applies to the messages you send in this chat.
The attach icon next to it adds an image to your message. It is only available when the selected model accepts images, and the tag under the input shows the upload status. The send button stays disabled while an upload is in progress, or if the file is too large or of an unsupported type.
Reading a response
The response streams into the chat. Steps the agent takes while working, such as searching items or reading an item, are grouped under a collapsible AI Reasoning section. The section stays open for the current response and is collapsed for earlier ones.
Proposals that need your approval
When the agent wants to change something it shows a card in the chat. Nothing is written until you accept it.
Create item - a card headed AI generated with the proposed content. Accept adds it to the bottom of the open document, and the arrow next to it lets you choose another position. See Item generation.
Edit item - a card showing the proposed attribute values, with a Highlight changes switch and an Apply button. See Reviewing and approving AI edits.
Create trace - a card headed AI suggestion showing the items and the trace type. Click Accept to create the trace or the bin icon to reject it.
Filter results - when the agent searches your items it shows the first results and an Expand results button, which opens the full result set as a new saved filter you can keep.
Saved analysis - when you ask for an analysis the agent builds it and shows an Open analysis button.
Stopping and resuming
While the agent is working, the send button becomes a stop button (Stop generating). A stopped chat shows Chat stopped. with a Resume button. If a response fails you see Chat response didn't complete. with a Retry button.
A response keeps running if you close the sidebar, navigate to another page, or close the browser tab. The chat shows Finishing your reply in the background... and the reply is there when you come back. Running chats are also listed under Agentic processes on the My feed dashboard, where you can stop or resume them.
Prompts that work well
"Find items that mention ISO 26262"
"Explain the term 'fail-operational' as used in this document"
"Summarize this document"
"Write 3 requirements about battery thermal limits"
"Generate a test case for @REQ-42"
"Rewrite the Description of @REQ-42 as a shall statement"
"Which requirements in this document have no verification trace?"
You can write in your own language and the agent answers in it.