How the assistant takes action across your apps

The assistant is more than a chat that answers questions. When you ask it to get something done, it finds the right capabilities across your connected apps and carries out the work for you — drafting documents, managing your tasks, sending messages, and more. This article explains how that works and how you stay in control.

From a request to real work

You describe what you want in plain language, and the assistant figures out how to do it. It selects the appropriate tools for the job — whether that means searching the web, pulling live data, creating a file, or reaching into an app you've connected — and runs them as part of its reply.

As it works, you see live status while each step runs, so a multi-step request feels transparent rather than a black box. You don't need to name a specific tool or command; describing the outcome you want is enough.

Tip: Be specific about the result you want. "Draft a follow-up email to Priya about the Q3 proposal and show it to me before sending" gives the assistant a clear goal and a clear stopping point.

What the assistant can do for you

Depending on which apps you've connected, the assistant can take a wide range of actions directly from the conversation. Common examples include:

• Draft and create documents, spreadsheets, images, and charts you can download or reuse
• Search the web and live data sources to research a question
• Manage your tasks, reminders, notes, contacts, and meetings
• Send email and messages through your connected apps
• Place phone calls on your behalf
• Create and manage your AI agents

The more apps you connect, the more the assistant can do without leaving the chat.

Connecting your apps

The assistant can only act in apps you've linked to your account. To expand what it can do:

1. Open your Integrations settings.
2. Find the app you want to connect and follow the prompts to authorize it.
3. Return to the chat and ask the assistant to do the work — it will use the newly connected app automatically.

If the assistant tells you it can't complete something, it's often because the relevant app isn't connected yet. Connecting it and asking again usually resolves this.

You choose when it checks with you

Some actions have real-world effects, such as sending a message, deleting something, or making a payment. Your permission mode decides when the assistant checks with you before these: pick Auto, Manual or Bypass permissions with the mode button under the chat box.

When it checks with you, a prompt appears above the chat box:

1. Review the details of exactly what the assistant is about to do.
2. Choose Allow once or Always allow to let it run, or Deny to stop it.
3. If something looks off, deny it and tell the assistant what to change.

Read-only work — like researching, drafting, or analyzing a file — generally runs without asking, since nothing is sent or changed. See Choosing when Praxivara asks first for how each mode works.

Confirming that an action happened

After the assistant completes a task, it summarizes what it did. If you ever want to double-check an outbound action, the confirmation lives in the app itself — for example, a sent email appears in your connected mail app's Sent folder, and a created task shows up in your task app.

Note: If a response was interrupted partway through a multi-step task, ask the assistant to continue or check the current state before repeating the request, so an action isn't run twice.

Staying in control

You steer the work at every point:

• Use the Stop button to interrupt a response that's heading in the wrong direction.
• Deny a permission prompt to stop an action before it runs, or switch to the mode that suits you.
• Give the assistant follow-up guidance — a change of format, tone, or recipient — and it will adjust.
• Disconnect an app from your Integrations settings at any time to remove the assistant's access to it.

Together, these controls let the assistant handle real work across your apps, the way you choose.