Graphnosis

Local, encrypted, deterministic job memory — for professionals and the agents you build

AI memory with zero network dependency

defense contractors · manufacturing OT teams · classified-facility IT · offline research labs · critical infrastructure operators

Graphnosis was built offline-first. In air-gapped environments, every feature — ingestion, recall, Skills, sharing — operates without an internet connection, an API key, or a cloud account.

Defense & classified facilities

Deploy on isolated networks or SCIF-adjacent infrastructure. No outbound traffic, no cloud dependency. Memory sharing via signed encrypted packs — no network relay ever.

Manufacturing & OT

Ingest SCADA historian data, OPC-UA feeds, runbooks, and maintenance logs. AI clients on the plant floor recall the right procedure without crossing the IT/OT boundary.

Offline research labs

Index papers, datasets, experimental notes, and protocols on an isolated research network. Local LLM (Ollama) powers semantic recall without external inference calls.

Edge & embedded hardware

Runs on Apple Silicon, NVIDIA Jetson, AMD, and Intel Arc. Low-overhead indexing designed for resource-constrained environments.

Why Graphnosis fits

  • Fully offline: indexing, recall, and Skills all work with $0 in LLM calls and zero network dependency.
  • Local LLM via Ollama: semantic recall and distillation on-device, no external inference API.
  • SCADA, OPC-UA, Modbus TCP, MQTT, and LoRaWAN data sources supported — no IT/OT boundary crossing.

Shared Engrams

GEZ — Graphnosis Engram Zero (.gez) is the offline memory-transfer format built into Graphnosis. The name reflects its design principle: zero network dependency — the file is the entire transfer mechanism.

A .gez pack is a self-contained encrypted bundle of one engram: every source, every memory node, ready to carry on a USB drive or push through a one-way data diode. It is AES-256-GCM encrypted and optionally signed with the sender's cortex Ed25519 key, so the recipient can verify authorship and integrity before importing — with no handshake, no cloud relay, no trust in the transport medium. export_engram creates the pack; import_engram ingests it on the other side, skipping sources that already exist. Signature verification happens offline against a public key the sender shares out-of-band.

Pro & Teams

Pro and Teams tiers add Autonomous Skills for SOPs and runbooks, Job Memory Kits for fast deployment, and the Memory Studio graph view — all of which work fully offline. Contact us for volume licensing and procurement support.

See plans →

Looking for the full picture — all use cases, how it works, AI clients, privacy details, and more?

See everything on the Graphnosis homepage →