Roadmap

Working today, on every supported platform.

• Cross-platform diagnostics

  macOS, Windows and Linux. Crash history, disk health and SMART, memory pressure, thermals, power and network reachability, using the tooling each OS already ships.

• Correlation engine

  Reads findings against each other rather than listing them. Three panics and a die near throttle are one problem, not two.

• On-device explanation

  A language model running on your own machine turns the findings into plain English. No API key, no account, nothing leaves the machine.

• Hardware-matched model sizing

  First run measures the machine and picks a model size it can run well, weighing free memory, cores and VRAM. Fitting is not the same as being fast enough to wait for.

• Bring your own model

  Point Triage at an Ollama install, a llama.cpp build, or anything that takes a prompt. The diagnostic prompt stays ours, so a connected model is held to the same instructions.

• Hardware escalation

  The agent pages your on-call by webhook when a physical fault appears. Slack incoming webhooks work directly. Software faults never page.

• Built for pipelines

  --json output, meaningful exit codes, and --offline for air-gapped runs that make no outbound connection at all.

In progress, targeted at the 25 September release.

• Getting a model onto the machine

  Triage knows which size suits your hardware. Deciding whether it fetches that model itself or hands you the exact one-line command is the open question.

• Licence activation service

  Exchanges a purchase key for a device-bound licence. The only server in the product, and it exists purely because one machine cannot know how many others a key is already on.

• Buying it

  Checkout, so a personal licence does not depend on emailing us first.

On the list, not in the build. Ask us for dates before you buy against any of these.

• Autonomous remediationEnterprise

  Fixing the reversible column on its own, with a dry run first, an audit log, and a policy saying what it may touch.

• Always onEnterprise

  Running as a background service so degradation is caught before it becomes an outage, rather than when somebody thinks to look.

• Custom severity policyEnterprise

  Your thresholds rather than ours. A 7% free disk is an emergency on a database host and a Tuesday on a laptop.

• Managed deploymentEnterprise

  Signed packages that install silently through Jamf, Intune, or your own MDM.

• Air-gapped model bundleEnterprise

  Runner and weights shipped alongside the binary, so there is no download and no egress to get approved.

• Fleet console

  One rollup across every enrolled device. The agent and its protocol exist; the console is not generally available yet.