Assistant 0.5 Release Notes

Release date: November 2026

Assistant 0.5 understands your graph. It answers questions about your environment from your own data, not a guess, and it combines what is in your account with how JupiterOne works. Every answer cites its sources, so you can check it.

Do I need to do anything?

No. Select Ask AI at the top of any page to start.

What is new in Assistant 0.5

Capability	What it means for you
Answers about your environment	Ask about your assets, identities, findings or integrations, and the assistant answers from your data
Explains your graph and data model	Learn what your account holds, where it comes from and how it connects, with links to see it
Your graph and JupiterOne together	Ask how something works, and get the mechanics and what they look like in your data
Shows its work	Every answer cites its sources, and you can see each step the assistant took

Answers about your environment

Ask about your environment in your own words. The assistant explores your graph, runs queries, reads the results, and refines its queries until it has the answer. Its answers are grounded in what is actually in your account.

• "Do our users have MFA enabled?"
• "Which hosts have no agent installed?"
• "What integrations do I have, and which ones are failing?"

The assistant sees only what your user account can see in JupiterOne.

Explains your graph and data model

Ask what your graph is made of, or how a class fits together, and the assistant describes your data: which classes and types you have and roughly how many, which integrations they come from, and how they relate. Classes, types and integrations in the answer link to assets inventory and the integrations page, so you can go straight to the data.

• "Tell me about the shape of my graph."
• "How do hosts fit together in my account?"
• "What does UnifiedDevice hold here?"

Your graph and JupiterOne together

The assistant answers from your data and from the JupiterOne documentation in the same answer. Ask how a feature works, and it explains the feature and what it looks like in your account.

• "How is risk scored, and how is it configured here?"
• "How does device unification work, and how many of my hosts are unified?"

Shows its work

Under each answer, N sources lists what the answer is based on. In the answer itself, a small label marks each statement with its source:

Label	What it means
Your data	Something in your graph, such as a count or a property
Query	A query the assistant ran to answer
Docs	A page of the JupiterOne documentation
JupiterOne	How JupiterOne itself works
On screen	Something shown on the page you are on

While it works, you see each step as it happens. When it finishes, the steps fold into one line, such as Worked for 8.3s · 2 steps; expand it to see each step, the queries it ran and how long each took. If a step fails, it stays visible, and the assistant can try another way.

Also in this release

• Links into the app. Entities in an answer open in the inspector; classes and types open in assets inventory; integrations open that integration's instances.
• Queue your next question. Messages you send while the assistant works wait under Queued. Select Send now to send one right away.
• Suggested questions. A new conversation starts with up to three questions picked for the page you are on and your data.
• What's new. Open the assistant as a full page, and its suggestions lead with what is new in JupiterOne.
• About cards in assets inventory. Each class and type has a short description and suggested filters.

Getting started

1. Select Ask AI at the top of any page
2. Ask about your environment in your own words, or select a suggested question
3. Open N sources to see what the answer is based on

For what came before, see What's new in Assistant 0.4. For the full list of JupiterOne AI capabilities, see JupiterOne AI capabilities.