PoisonZero Pro gives teams real-time monitoring and automatic rollback of poisoned memory entries - on up to 15 devices, with a Cloud or Private install, starting from the very first device.
The poison0 agent watches your agents' memory files and reverts poisoned entries before they influence the agent.
Every decision is logged in the console with time, danger score and outcome.
Set thresholds, protected paths and profiles from the console, with no config files to edit.
Run cloud-managed or fully private with a zero-egress option, at no extra cost.
One install flow with OS-native paths and service manager.
Melira is trained for one job, spotting poisoned memory writes, in many languages.
Your verdict on an incident can become a training candidate, only after an independent re-check; nothing trains automatically.
The engine reads only the model file, answers only on localhost and spawns no processes.
If the engine crashes, hangs or answers oddly, the change is recorded and surfaced, never waved through.
The on-device detection engine runs on the hardware your team already has - quietly, on demand, without a GPU. These specs apply to Private-mode devices; Cloud-mode devices send only redacted diffs for scoring and have a lower local footprint.
| Property | Detail |
|---|---|
| Footprint (Private mode) | A little over 300 MB - and only for a few seconds during a memory check. At rest, just a few MB. |
| Hardware | CPU-only, ordinary hardware - no GPU required. GPU acceleration is an optional, opt-in add-on. |
| Analysis latency | A few seconds per memory check, started on demand. |
| Network footprint (Private) | One optional request per month - credentials and version only. Cloud-mode devices also send redacted diffs; that egress is documented in the whitepaper. |
| Updates | Signed artifacts, SHA-256-verified before every start. |
GPU acceleration: what changes
Create an app in the console, connect your device with an enrollment code - and your first Memory Files are guarded.
Sign me up