One box. The whole system.
Storage, database and compute, preconfigured. Rack it, provision your workspace, and your team syncs, searches and asks — with every byte staying behind your own firewall. A true on-premises AI platform — no cloud, no egress, no compromise. The same discipline as the cloud edition: labels on landing, walls on every read, metered AI, human-approved actions.
That includes the AI boundary: point Memkist Local at your own model endpoint and usage is still metered per purpose and visible in the local console. "On-prem" is not a license to fly blind.
Spec at a glance
| Deployment | On-premises appliance |
| Data residency | 100% local · no cloud sync |
| Platform | Sync · Search · AI assistant · Workflows · AI tools |
| AI model | Private endpoint on your network |
| Access | Desktop · Web · connect your AI apps |
| Availability | Coming soon |
Everything with your name on it.
Fully on-premises
Every byte — storage, parsing, search and agent runs — stays on the appliance, behind your own firewall. Fully air-gapped deployments supported. No egress, no third party in the data path.
Same agent platform
The AI assistant, memory, workflow compiler and MCP server run locally — the same behaviour as the cloud edition, on your hardware.
Private model endpoint
Point the platform at an in-house LLM on your network. AI usage stays metered and auditable, entirely on your infrastructure.
Drop-in appliance
A single box with storage, database and compute preconfigured. Rack it, provision your workspace, start working.
You don't have to wait.
The cloud edition is the same platform, available today — its per-seat pricing is being finalised and will be published soon. Start in the cloud, move to the appliance when it ships — no rebuild, no re-training, same platform.
Questions about Local?
Residency requirements, hardware sizing, air-gapped deployments, your own model endpoint — write to us and get answers from the people building it.