Now accepting early access developers

Prove you
destroyed the data.

Process sensitive data in a sealed ephemeral environment. Get a cryptographic proof it was destroyed. Verify it without trusting us.

Free tier · 100 sessions/month · No credit card

nanorix — ephemeral session
Use Cases
Built for compliance-heavy industries.
🏥

Healthcare & HIPAA

Process patient data for diagnostics or analytics — then prove to auditors every byte was destroyed. Replace compliance consultants with cryptographic proof.

HIPAA § 164.530(j) · Destruction verification
⚖️

Legal & Privilege

Analyze privileged documents through AI tools without creating persistent copies. Attorney-client privilege requires proof of non-retention.

ABA Model Rule 1.6 · Document lifecycle
🏦

Financial Services

SOC 2 requires documenting data lifecycle management. Nanorix generates audit-ready destruction receipts automatically — no manual documentation.

SOC2 Type II · CC6.5 data disposal
🤖

AI/ML Training

Train models on licensed datasets where the license requires proof of deletion after use. Process the data, train the model, destroy the data, prove it.

Data licensing compliance · Right to be forgotten
The Problem
Your data outlives your intent.

You process sensitive data — patient records, legal documents, financial models. When you're done, you delete it. But "deleted" doesn't mean gone. And you can't prove otherwise.

🔓

Deletion is a lie

Filesystem journals, swap space, kernel logs, and memory residues survive "deletion." Forensic tools reconstruct what you thought was gone.

📋

Compliance is theater

You document data handling in spreadsheets. Auditors ask for proof of destruction. You produce screenshots and promises. Nobody is convinced.

🧩

Privacy is bolted on

VPNs, encrypted disks, access controls — all layered on top of systems designed to remember everything. The foundation fights the goal.

Nanorix replaces trust with cryptographic proof.

How Nanorix Works
Four API calls. Zero residue.

Every computation happens inside a cryptographically sealed session that self-destructs. No logs survive. No data persists. The proof is mathematical.

verify.py
from nanorix import NanorixClient

client = NanorixClient(api_key="nanorix_sk_live_...")

# 1. Create a sealed ephemeral session
session = client.create_session(ttl_seconds=300)

# 2. Process sensitive data inside the sandbox
result = session.execute("analyze(patient_records)")

# 3. Destroy everything — get your cryptographic proof
cdp = session.destroy()

# 4. Verify the proof — anyone can, no trust required
assert client.verify(cdp).valid
1

Create Session

Sealed sandbox

POST /v1/sessions
2

Send Data

Volatile memory

POST /v1/sessions/{id}/exec
3

Destroy Everything

12-step wipe

DELETE /v1/sessions/{id}
4

Receive Proof

Math proves it

POST /v1/verify

Your data enters volatile memory. Results come out. Everything else is destroyed through a 12-step cryptographic sequence, signed with an ephemeral Ed25519 key — which is itself destroyed after signing. The proof is yours forever. The data is gone forever.

Integration
Fits your existing pipeline.

Add ~10 lines to your backend. Choose the pattern that matches your workflow.

Best for: new pipelines or teams migrating to zero-PHI architecture

Process inside Nanorix. Raw data never reaches your servers.

Send data and computation to Nanorix. Your code runs inside our volatile sandbox. Only processed results come back. Raw data is destroyed with proof.

Data source
Nanorix processes
Results only + CDP
Your servers never hold raw sensitive data. Smaller HIPAA scope. Simpler audits. Automatic destruction proof for every record.
pipeline.py
# Hospital sends patient data to your API
patient_data = request.json["record"]

# Forward directly into Nanorix — never store locally
session = nanorix.create_session(ttl_seconds=300)
result = session.execute(
    f"python model.py --data '{patient_data}'"
)

# Only the diagnosis leaves the sandbox
diagnosis = result.output

# Raw PHI destroyed — proof generated
cdp = session.destroy()
cdp.save(f"proofs/{session.id}.json")

# patient_data was never written to your disk,
# your database, or your logs.
return {"diagnosis": diagnosis}
Best for: API-driven architectures and webhook receivers

Nanorix is the first and only thing that touches the data.

Point your data source directly at Nanorix. Incoming records land in volatile memory, get processed, and only derived results reach your system.

Hospital API
Nanorix receives + processes
Your system gets results
Raw data never exists outside Nanorix. No copies on your servers. No copies in your caches. No copies in your logs.
gateway.py
# Webhook: hospital sends records directly
@app.route("/intake", methods=["POST"])
def receive_patient_data():
    raw = request.data  # don't parse, don't store

    # Immediately forward into Nanorix
    session = nanorix.create_session(ttl_seconds=120)
    result = session.execute(
        "python classify.py", stdin=raw
    )

    # Raw PHI destroyed with proof
    cdp = session.destroy()
    cdp.save(f"proofs/{session.id}.json")

    # Your DB stores the score, never the record
    db.insert(score=result.output, proof=cdp.root_hash)
    return {"score": result.output}
Best for: enforcing data retention policies on existing records

Prove you destroyed what you already have.

Historical data that needs to expire — retention policies, patient requests, regulatory purges. Send each batch through verified destruction. Get a CDP for your audit trail.

Expired records
Verified destruction
CDP per batch
Automated audit trail. Each purge cycle generates verifiable proof. Auditors get a folder of CDPs — not a spreadsheet.
compliance_job.py
# Monthly: destroy expired patient records
expired = db.query(
    "SELECT * FROM patients WHERE expires < NOW()"
)

for batch in chunk(expired, size=100):
    session = nanorix.create_session(ttl_seconds=60)
    session.execute(
        f"echo '{json.dumps(batch)}' > /tmp/data"
    )
    cdp = session.destroy()
    cdp.save(f"compliance/{cdp.cdp_id}.json")

    # Now safe to delete from your database
    db.execute("DELETE FROM patients WHERE id IN %s",
               [r.id for r in batch])

# Auditor: nanorix verify compliance/*.json
Best for: long-lived workloads — months or years before destruction

Install once. Destroy when ready. Same proof.

Some sessions start in 2026 and end in 2028. A lightweight Nanorix agent runs on your infrastructure — your GPUs, your storage, your schedule. When the retention policy expires or you decide to destroy, the agent runs the full 12-step destruction chain locally and generates a CDP.

Your infrastructure
Nanorix agent destroys
CDP generated locally
Same 12-step chain. Same CDP format. Same offline verification. Sessions can run for days, months, or years. The agent waits until you're ready.
terminal
# Install the Nanorix agent
$ curl -sSL https://install.nanorix.io | sh

# Configure with your API key
$ nanorix-agent init --api-key nanorix_sk_live_...

# Set a retention policy — destroy after 90 days
$ nanorix-agent watch /data/patients/ --destroy-after 90d

# Or destroy on demand — any time you're ready
$ nanorix-agent destroy /data/case-2024/ --generate-cdp

✓ 12-step destruction chain complete
⬡ CDP saved: proofs/cdp_a8f3e71c.json
  Verify: nanorix verify proofs/cdp_a8f3e71c.json
Request Early Access →

Agent beta launching Q3 2026

The Paradigm Shift
Privacy as architecture, not afterthought.
Current Approach

Privacy as Policy

  • Add encryption after building the system
  • Write policies about data handling
  • Trust users to follow deletion procedures
  • Hope auditors accept your documentation
  • Data persists by default, deleted by effort
$50 – $500
per manual destruction certificate
vs
Nanorix Approach

Privacy as Architecture

  • Data is ephemeral by construction
  • Destruction is automatic and verified
  • Proof is cryptographic and mathematical
  • Nothing persists by default, stored by exception
  • Compliance proof: one API call
$0.02 – $0.05
per cryptographic destruction proof

1,000× to 2,500× cheaper — using the lowest manual certificate cost ($50). Most enterprises pay $150–$500, making the real savings 3,000× to 25,000×.

The Proof
Don't trust us. Verify it yourself.
destruction_proof.json
{
  "version": "2.0",
  "session_id": "sess_a8f3e71c...",
  "terminated_at": "2026-02-27T14:30:00Z",
  "proof_chain": {
    "steps": [
      { "step": 0, "subsystem": "eee_namespace",
        "hash": "7d4f8a91b3c7e2..." },
      // ... 10 more steps, each hash-chained
      { "step": 11, "subsystem": "session_state",
        "hash": "2c8a4f63d19e..." }
    ],
    "proof_hash": "2c8a4f63d19e..."
  },
  "root_hash": "91d7e580a4b2...",
  "attestation": {
    "algorithm": "Ed25519",
    "signature": "VGhpcyBpcyBhIG...",
    "public_key": "MCowBQYDK2Vw...",
    "key_destroyed": true
  }
}

12-Step Hash Chain

Each step hash-chains to the previous. Tamper with one step and the entire chain breaks.

Root Hash

Binds destruction chain, session ID, and timestamp into a single verifiable hash.

Ed25519 Attestation

Ephemeral key signs the root hash, then destroys itself. Mathematically bound to the destruction event.

Key Destruction

The signing key is zeroized immediately after use. One key, one proof, forever.

Self-Contained Trust

The CDP is the trust anchor — not our database. Store it. Verify it in 5 years. No dependency on us.

Verify with 6 lines. No network. No trust.
verify_offline.py
from nanorix.verify import verify_cdp_offline
import json

with open("destruction_proof.json") as f:
    cdp = json.load(f)

result = verify_cdp_offline(cdp)
print(result)  # ✅ Chain valid · Signature valid · Root hash valid

The verification algorithm is public. The Python SDK is open source. Your auditor can run this themselves.
Full CDP specification →

Security
Five layers. Zero persistence.
05
Ed25519 ephemeral signingOne key per session. Destroyed after single use. CDP mathematically bound to destruction.
04
cgroup v2 resource limitsPer-session memory caps. No resource escape. OOM kills contained.
03
seccomp-bpf syscall filtering~70 allowed syscalls. Everything else kills the process. No exceptions.
02
Volatile memory onlytmpfs with noexec. Swap disabled. No data touches disk. Ever.
01
Linux namespace isolationUser, mount, PID, UTS, network. Each session is a sealed universe.

Written in Rust. No garbage collector. No runtime surprises.

Read our security model — including what we don't protect against →
The Chain
12 steps. Two layers.

Steps 0–2 destroy your data. Steps 3–11 destroy every trace the system held about it.

Customer Data Destruction
0eee_namespaceLinux namespace teardown
1eee_tmpfsFilesystem unmount + destruction
2eee_memoryMulti-pass memory overwrite

After step 2, your data no longer exists in any form.

System Trace Elimination
3dire_keysCryptographic key zeroization
4dire_identitySession identity destruction
5rzl_auditAudit trail purge
6ace_complianceCompliance context purge
7fgi_forensicForensic identity destruction
8zilchnetNetwork state purge
9memproofMemory protection state purge
10sis_pipelinePipeline isolation teardown
11session_stateFinal metadata purge

System state can reveal information about the data it processed. We destroy that too.

Pricing
Simple, transparent pricing.

Every tier includes signed CDPs, public verification, offline verifiability, Python SDK, and CLI verifier.

Free
$0/mo
For evaluation and development
  • 100 sessions/month
  • 60-second max TTL
  • 3 concurrent sessions
  • 20 requests/min
  • Signed CDPs included
Starter
$99/mo
$0.05/proof
For production workloads
  • 2,000 sessions/month
  • 30-minute max TTL
  • 20 concurrent sessions
  • 120 requests/min
  • Priority support
Most Popular
Business
$499/mo
$0.02/proof
For compliance-critical operations
  • 25,000 sessions/month
  • 2-hour max TTL
  • 100 concurrent sessions
  • 300 requests/min
  • Dedicated support
Enterprise
Custom
For regulated industries at scale
  • Unlimited sessions
  • Custom TTL + concurrency
  • SLA guarantees
  • Architecture review
  • Compliance mapping

Each CDP replaces a $50–$500 manual destruction certificate. That's 1,000× cheaper.

All plans scale. Overages billed at tier rate. No hidden fees. Cancel anytime.
FAQ
Questions developers ask.

Deletion marks space as available — it doesn't erase data. Filesystem journals, swap space, kernel logs, and memory residues all retain fragments. Nanorix never writes to persistent storage. Data exists only in volatile memory and is cryptographically shredded at termination. The destruction proof verifies this mathematically.

A 12-step hash-chained receipt covering every destruction component: namespace teardown, tmpfs unmount, multi-pass memory overwrite, key zeroization, identity destruction, and zero-state verification. Each step is independently verified and hash-chained so tampering is detectable. Signed with an ephemeral Ed25519 key that's destroyed after signing.

Nanorix provides what auditors actually want: verifiable proof that sensitive data was destroyed after processing. The destruction receipt is a cryptographic artifact — not a screenshot, not a policy document, not a promise. It's mathematical evidence with an independently verifiable hash chain and Ed25519 signature.

Phase 1 is a cloud API for workloads that fit inside ephemeral sessions (up to 2 hours). For GPU-heavy workloads or long-lived data, our roadmap includes an embeddable agent that runs on your infrastructure — your GPUs, your storage, your timeline. When you're ready to destroy, the agent handles verified destruction and generates CDPs locally. Same proof, your hardware.

Yes. The verification algorithm, CDP specification, Python SDK verify_cdp_offline() function, and CLI verifier are all open. Security researchers and your auditors can independently verify destruction proofs without trusting our infrastructure. The core engine is patent-pending.

Session creation takes milliseconds. Execution runs at native speed inside Linux namespaces — no virtualization overhead. Destruction adds a small overhead for multi-pass memory wiping and proof generation, but the entire lifecycle completes in under a second for typical workloads.

Get Started
Start verifying destruction.
The Vision
Nanorix is bigger than CDPs.

Cryptographic destruction proofs are the beginning. The Nanorix architecture covers ephemeral execution, disposable identity rotation, sealed application isolation, zero-knowledge networking, and adaptive privacy governance. We're building the infrastructure layer for identity-free computing.

Patent pending · 133 claims · 12 subsystems

Patent Pending · Ed25519 · Rust · Open Verification Spec · HIPAA · GDPR · SOC 2