Local AI
64 GB or more is recommended for the local path.
YOUR RTX SPARK STARTS HERE
Unlock next-level AI on your device with guided playbooks for agents, local models, creation, and development.
Turn your Spark into a private AI teammate for email, calendar, and file management.
Open playbook →FIRST TIME HERE?
GUIDED AI WORKFLOWS
Step-by-step guidance built around what you want to accomplish.
First 15 planned playbooks from the working roadmap.
RTX SPARK PLAYBOOK · 15 MIN
Turn your PC into a private AI teammate for the work that fills your day.
01
Turn your RTX Spark into a personal AI teammate that can help with email, calendar, and file management. Add focused tools and skills, and it becomes a chief of staff that can organize information, prepare next steps, and keep routine work moving—with you approving consequential actions.
Run supported models locally for a responsive, private experience without per-token API costs, or use a cloud model when you need more capability. You stay in control of what the assistant can access and when it can act.
02
64 GB or more is recommended for the local path.
All RTX Spark configurations can use a supported cloud model.
Have only the work accounts you want to connect ready to authorize.
Start with read-only access and require confirmation for messages, file changes, calendar edits, sharing, purchases, and other consequential actions.
Select the experience you want. The install button hands off directly to the app installer.
Open your selected app and follow its setup. Choose Local if your RTX Spark has 64 GB or more. Otherwise choose a supported cloud model.
Copy the setup prompt from this page and paste it into the installed agent. It will inspect your setup, find the official NVIDIA skill package, and ask before downloading or changing anything.
Tools let the agent interact with an app or service. Skills give it a proven sequence for a specific job. Together, they turn a general assistant into a focused chief of staff.
Approve the official NVIDIA skill package only after the app verifies its source. Review the permissions it requests before enabling it.
Connect only what you need: email, calendar, Google Drive, Slack, and Microsoft Teams. Start with read-only access, then expand permissions one service at a time.
Ask: “Review today’s messages and calendar, then give me a private briefing with three priorities.” Check that no external action happens without approval.
Require approval before the assistant sends messages, changes meetings, moves files, shares data, or makes purchases. Ask it to explain when a task leaves the device for cloud processing.
03
Try these quick checks before restarting setup.
Close memory-heavy apps, confirm at least 64 GB for the local path, and choose a smaller supported model. If local inference still cannot start, switch to cloud.
Reconnect the service, confirm the correct work account, and approve the minimum permissions. Company policy may require administrator approval.
Confirm the skill came from the official NVIDIA source, restart the app, and check Skills or Extensions. Do not install similarly named files from unverified sources.
Disconnect the tool, reset it to read-only, and add an explicit confirmation rule for messages, calendar changes, files, sharing, purchases, and device control.
Paste it into a new chat after onboarding. If the app cannot install software, use the prompt as a checklist and approve each step manually.