Every product. One engine, one policy.
Data reaches a language model three ways: people type it, agents pass it to tools, teams upload documents. Each Nyx product covers one path, and all of them share the same privacy engine, the same signed policy and the same audit trail.
Say yes to AI chat, without the leak.
Your teams keep the assistants they already use. Nyx rewrites each prompt and upload on the device so it no longer identifies anyone, removes secrets, and stops requests that should never be sent.
Watch the live demoNyxEndpoint
- Covers ChatGPT, Claude, Gemini, Copilot and more, in the browser and in their desktop apps
- Personal details generalized by meaning, secrets and credentials removed
- Anonymity measured across the whole conversation, not one message at a time
- Uploaded files, scans included, cleaned before they reach the provider
- A developer mode that keeps code useful while secrets still stay home
- Reports its own health, so a bypassed device shows up in the console
NyxBrowser
- For devices where a desktop agent can't be installed
- The same privacy engine, running inside the browser
- Works alongside NyxEndpoint on the same machine without double-processing
Give agents tools, not your data.
NyxGate is the security gateway between AI agents and their MCP tools. It checks what tools tell your agents and what your agents send to tools, and stops the call that turns a hidden instruction into a leak.
Explore NyxGateNyxGate
- One address per MCP server, set up from the console, no agent code changes
- Stops poisoned tools, tools that change after approval and lookalike servers
- Blocks data leaving through a tool when the instruction came from untrusted content
- Asks the user before destructive actions: dropped tables, force-pushes, wiped folders
- Personal data in tool arguments swapped for placeholders and restored in the answer
- Upstream credentials held by the gateway, never in developers' configs
- A signed receipt for every session, and a hash-chained activity log
Send the analysis, not the people.
NyxCloak anonymizes a whole set of documents as one: a roster and a case note that each look safe but identify someone together are caught and fixed consistently across every file.
Open NyxCloakNyxCloak
- Whole folders of PDFs, Office files, spreadsheets and scans in one job
- Identity generalized, the facts the analysis needs kept intact
- Every set checked against a k-anonymity target before it's released
- A signed receipt stating how anonymous the set is
- Web console, REST API and Python and TypeScript SDKs
One console. One policy. One record.
NyxCommand governs the whole suite. Write a policy once and every product enforces it; see people, agents and documents in one place; hand auditors evidence built from what actually happened.
See the compliance evidenceNyxCommand
- Policies signed and pushed to every product in the suite
- Fleet health, incidents with forensic detail and threat mapping
- NyxGate servers, access tokens and live gateway activity
- Tamper-evident, hash-chained ledger of every decision
- Evidence packs for the EU AI Act, GDPR, NIS2, DORA, ISO 27001, ISO 42001, SOC 2, HIPAA and PCI-DSS
- Human review of automated decisions (GDPR Art. 22)
One engine underneath
Semantic k-anonymity
Patent-pending generalization backed by real population data, so "anonymous" is measured, not assumed.
On-device analysis
Prompts are classified locally. Nothing is sent to a third party to decide whether it was safe to send.
Signed, live policy
Policies and coverage updates are signed and pushed in minutes, including new AI provider endpoints.
Proof by default
Hash-chained ledger and signed receipts, verifiable without trusting Nyx.
More that sets Nyx apart
Capabilities that ship today, and the parts most AI-security tools leave uncovered.
Self-updating coverage
When an AI provider changes or adds an endpoint, Nyx detects it from fleet telemetry and extends protection in minutes, not on your next release cycle.
Both directions
Most tools only watch what goes out. Nyx also inspects what comes back: model replies for injection echoes and manipulation, tool results for hidden instructions.
Document inspection
Uploads aren't a blind spot. PDFs, Office files, spreadsheets and scans are read, checked for embedded threats and anonymized with the prompt they came with.
Defense in depth
Desktop, browser, agent tools and the network edge, layered so a request that slips past one control still meets the next.
Encrypted at rest
Forensic data and sensitive originals are sealed on disk with AES-256 and independently managed keys.
Fails closed
If a component degrades, Nyx defaults to protecting you. No silent bypass that quietly lets data through.
Runs where your data is allowed to live
SaaS, private cloud, on-premise or fully air-gapped, no telemetry leaves your enclave in sovereign deployments.
Nyx is deployed with your team.
There's no self-serve sign-up. We scope every deployment, cloud, on-premise or air-gapped, with you. Get a demo or request the technical whitepaper and we'll respond within one business day.