Skip to Content
FAQ

FAQ

Is my data private?

Yes. When using the desktop app, Alma processes your data locally on your machine by default. Cloud processing is opt-in on a per-source basis — you choose which sources, if any, are processed in the cloud. Your data is never shared between organizations.

What does Alma do with my emails and messages?

Alma extracts insights from your conversations — identifying people, companies, relationships, deals, and action items. It builds an organizational memory from these insights so you and your agents can reference them later. Alma does not store the full text of your emails or messages.

Can agents send messages without my approval?

Not unless you configure them to. Every tool and output channel on a responsibility can have an approval gate that requires a specific person or any team member to review and approve before the agent acts. You have full control over what runs automatically and what requires human review.

Do all team members need to connect their own accounts?

It depends on the integration. Personal connections like Gmail, Slack, and Google Calendar require each team member to connect their own account — Alma only accesses what each individual authorizes. Organization connections like GitHub, Linear, and Attio are set up once by an admin and shared across the workspace.

What happens if I disconnect a source?

Alma stops fetching new data from that service immediately. However, knowledge that was already extracted remains in your organizational memory. If you want to remove that knowledge, you can do so from the Memory section.

Can I use Alma without the desktop app?

Yes. The web app at alma.team  works on its own. If you want your sources to continue syncing without the desktop app running, enable cloud processing for each source in Settings → Connections.

How do I control what an agent can do?

Each agent has a defined set of tools (integrations it can use), skills (instruction sets it can follow), and responsibilities (automated tasks). Within each responsibility, you can configure approval gates on individual tools and output channels. This gives you granular control — for example, an agent might be able to read your email automatically but require approval before sending a reply.

What AI models do agents use?

You choose the model when creating an agent. Alma supports multiple AI providers, so you can select the model that best fits each agent’s tasks.

Last updated on