Trust · Responsible AI by design
Letting AI act for your company takes more than a good model. Here is exactly what Constella can see, what it can do, when it asks, and how it gets better.
You set goals and limits. Accountability stays with people.
Every answer cites the source it came from, so anyone can check it.
Change models or vendors any time without losing what you know.
01 · Your data
| Plan | Where it's stored | Models | Used to train AI models? |
|---|---|---|---|
| Free | Secure cloud, private to your account | Leading AI providers, under zero data retention agreements | No. Providers don't keep your data or train on it. |
| Team | Secure cloud, scoped to your team | Leading AI providers, under zero data retention agreements | No. Providers don't keep your data or train on it. |
| Enterprise | Your cloud or on-premises option | Self-hosted, your choice | Only your own model, for your advantage. Data and weights stay inside your walls. |
02 · Access
Your own captures are private to you until you add them to a project.
LIVEMemory is scoped by team, project and role. Agents only reason over what's in scope.
LIVERemove a member and their access ends immediately. What they captured stays with the company.
LIVEDecide what's remembered, for how long, and who can use it.
ENTERPRISE03 · Agents
On their own
Only after a person approves · COMING SOON
Never
Every agent has a plain-English Skill that sets its role, expertise and guardrails, and those guardrails ride along with every conversation. An activity record of what each agent read and did is coming soon.
04 · Learning
05 · Security
Constella is built and run to industry-standard security practices, from how data is stored and accessed to how changes reach production. Your security team can ask us for the details.
Constella is built on Silverberry's Responsible AI by Design framework: useful, explainable, and grounded in context people can inspect.
Not by anyone else. On our cloud plans, Constella works with AI providers under zero data retention agreements: under their terms, your data isn't stored by them or used to train their models. If you run a self-hosted model, your data can be used to fine-tune that model, inside your walls, so it gets better at your business and only yours.
Today, agents answer, summarize and draft from your sources. Approval rules for actions like sending or changing records are coming soon, and you will choose which actions need a person to sign off.
Remove them and their access ends immediately. The knowledge they added to shared projects stays with the company.
Yes. Your memory, rules and agents are kept separate from the model, so you can change models without starting over.