The Full Engine.
On Your Hardware.
CogmemAi Edge puts the complete Intelligence Engine, the same one behind the top published LongMemEval score, inside your own perimeter. Air-gapped networks, sovereign data requirements, robot fleets in the field: your Ai remembers everything, and nothing leaves the building.
Built for Environments Where the Cloud Is Not an Option
Some memory cannot ride on someone else's servers. Edge exists for exactly those cases.
Defense & Government
Classified and controlled environments with no external network path. Edge runs fully air-gapped: mission knowledge, operator corrections, and institutional memory stay inside the enclave, on dedicated equipment you control.
Robotics & Autonomous OEMs
Robot fleets, autonomous vehicles, and drones that operate with intermittent or zero connectivity. Edge gives every unit low-latency local recall, and what one unit learns can propagate to the fleet on your schedule, under your policies.
Regulated Industries
Healthcare, finance, and legal teams with residency, audit, and custody obligations. Edge keeps memory inside your compliance boundary while your Ai tools keep their full recall.
What an Edge Deployment Includes
This is not a stripped-down local cache. It is the production engine, licensed onto your infrastructure.
The Complete Intelligence Engine
Semantic extraction, precision reranking, the auto-linking knowledge graph, contradiction detection, consolidation, and behavioral auto-skills: the full stack that produced our published benchmark results, not a subset.
Quantum-Safe Encryption
Memory content is encrypted at rest with post-quantum cryptography. Keys are generated and held inside your perimeter; we never possess them.
Air-Gap Operation
Every function works without an outbound connection. Model updates and engine upgrades arrive as signed offline packages on the cadence your security team approves.
MCP + REST, Same as Cloud
The same MCP tools and REST API your teams already use against our cloud. Point existing clients at your Edge endpoint and nothing else changes.
Dedicated Engineering Support
Direct access to the engineers who build the engine: deployment assistance, tuning for your workload, and a named contact rather than a ticket queue.
License Terms Your Counsel Will Recognize
A conventional enterprise software license: annual term, named deployment scope, source escrow available, and audit rights defined on paper before anything ships.
Why Edge Is Never a Public Download
We publish our client tooling openly: the MCP server is on npm, the SDKs are on PyPI and npm, and the free tier requires no card. The engine itself is different. It embodies years of benchmark work, and our customers in defense and regulated industries require a controlled chain of custody for every binary inside their perimeter. A public image would serve neither purpose.
So Edge ships one way: to named customers, under contract, with license terms and support obligations both sides have signed. If that is the posture your environment requires, you are exactly who this page is for, and the first conversation costs nothing.
How an Edge Engagement Works
Scoping call, NDA if you need one
Thirty minutes on your environment: connectivity, hardware, compliance regime, and what your Ai systems need to remember. We will tell you plainly whether Edge fits or whether our cloud or hybrid mode serves you better at a fraction of the cost.
Pilot on dedicated equipment
A bounded pilot deployment against your real workload, with success criteria agreed in writing before it starts.
License and deploy
Annual license, named scope, signed offline update channel. Deployment is performed with your team, on your hardware, inside your change-control process.
Ongoing support
A named engineering contact, tuning reviews, and signed upgrade packages on your approved cadence.
Pricing
Edge is priced as an annual license scoped to your deployment. Most teams do not need it: our cloud tiers start free and hybrid mode covers most sovereignty concerns.
Edge (On-Premises)
- Full Intelligence Engine, on your hardware
- Air-gap and sovereign deployment
- Quantum-safe encryption, your keys
- Dedicated engineering support
- Annual license, defined scope
Cloud tiers from free to $99.99/mo are on the developer page.
Edge Questions, Answered
Is Edge the same engine as your cloud?
Yes. Edge is the production Intelligence Engine, the one behind our published 95.10% LongMemEval result, licensed onto infrastructure you control. Capability parity is a contractual commitment, not a marketing line.
Can it run fully air-gapped?
Yes. Every function operates without an outbound connection. Updates arrive as signed offline packages that your security team applies on its own schedule.
Who holds the encryption keys?
You do. Keys are generated inside your perimeter and never transmitted to us. We could not read your memory store even if compelled.
Do we need new client software?
No. The same MCP server and SDKs that talk to our cloud talk to your Edge endpoint. Existing agents, editors, and robots repoint with a configuration change.
Is there a downloadable trial?
No. Edge ships only to named customers under contract. For evaluation, the free cloud tier exercises the same engine and the pilot phase of an engagement runs on your own equipment.
What does it cost?
An annual license scoped to your deployment: unit count, sites, and support level. Scoping calls are free, and if cloud or hybrid mode fits your requirements we will say so and save you the money.
Explore the rest of CogmemAi: Developers & Pricing · Compare Ai Memory · Memory for Robots · Memory for Agents