Less time. More answers.
Built for IT teams with more devices than hands. Tell it what a user noticed, and it gathers the evidence, finds the likely cause and suggests what to do. Your team makes the call, and nothing runs until you approve it.
Everything on the grid runs in your environment. Only the evidence an investigation needs goes to your AI provider.
- Your devices run an agent that collects evidence and connects out over HTTPS.
- Your Hub is self-hosted and keeps the evidence.
- Your AI provider gets only what an investigation needs, with your key.
- An administrator approves every command before it runs.
How it works
Three steps, from a user’s complaint to an answer you can act on.
1Describe the problem
In plain words, or let it start from a user’s report, a security alert or a reading that’s unusual for that device.
2It gathers the evidence
Events, system state and history from the device, read-only, with facts kept separate from guesses.
3You approve the fix
You get a clear conclusion and next steps. If you allow it, the AI proposes the fix, waits for your approval, then checks that it worked.
What you get today
Everything in the current release →An AI chat for every device
Ask anything about a machine. It runs read-only checks when it needs fresh data and shows the evidence for its answer.
Fixes you approve
Choose Advise, Diagnose or Resolve. Fixes wait for approval, get checked afterwards, and the ones that work are remembered.
AI investigations
Likely causes ranked, with what supports each one and what is still unknown.
A help app for your users
People ask for help from their own Windows, Mac or Linux computer. Simple questions get answered, the rest reach your team with evidence attached.
Spots what’s unusual
Agents learn each device’s normal and flag readings far outside it, so problems surface before the ticket.
Security alerts, investigated
Alerts from SentinelOne, CrowdStrike, Defender or any webhook start a read-only investigation right away.
Safe by design
Powerful enough to answer hard questions, never powerful enough to get around your controls.
How we keep it safe →Read-only by default
Investigations only read. The AI picks from a fixed set of read-only tools and cannot change a device on its own.
Nothing changes without approval
Fixes, scripts and AI suggestions wait for an administrator, and you can require a second one.
Everything is logged
Roles for viewers, technicians and administrators, and an audit log of every sensitive action.
Chats stay on topic
Guardrails keep the AI to work and IT questions, for technicians and end users alike.
Runs on your infrastructure
Host the Hub yourself, use your own AI key, and keep the evidence in your environment.
Windows, macOS, Linux
Native agents, installed with one command and updated from the console.
No inbound ports
Agents connect out to your Hub over HTTPS.
Your AI key
Save OpenAI, Claude, Gemini or OpenRouter keys and pick a provider per chat. What will it cost?
Docker in minutes
Pull our published image and start it with Docker Compose.
Try it on your own devices
Join early access for the self-hosted preview. Pricing is being shaped with early users. See pricing