No. E2B, Modal, and Daytona are managed cloud services — your agent code runs on their infrastructure, your data leaves your environment. NeuronEdge Enclave is self-hosted — you run it in your own VPC or on-prem. The isolation technology is similar (Firecracker microVMs), but the deployment model and trust boundary are fundamentally different.

Additionally, NeuronEdge Enclave offers confidential mode (SEV-SNP hardware attestation + memory encryption). None of E2B, Modal, or Daytona offer hardware-attested, operator-excluded execution (as of mid-2026).

Standard tier: each workspace is a Firecracker microVM with its own kernel. The host operator is trusted (no memory encryption). This is the multi-tenant / untrusted-code / cost-efficient path.

Confidential tier: the host itself is a SEV-SNP confidential VM. The workspace runs directly inside it, memory-encrypted by hardware. The cloud operator cannot read the agent's memory. Key release is gated on hardware attestation. One CVM per sensitive workspace.

Both tiers share the same API, SDKs, and audit surface. Switch with a single env var: NE_CONFIDENTIAL_MODE=1.

The audit chain is a sequence of Ed25519-signed events forming a hash chain. Any party holding an exported manifest can independently verify every signature, every chain link, and detect tail truncation — without trusting the host. nee audit export writes the verifiable artifact; nee audit verify confirms it.

Honest caveat: a root operator who deletes the local log before any export can destroy evidence. Automated periodic export + a WORM sink are the operator's responsibility.

2026 estimates based on published CSP benchmarks. We publish measured percentiles per host SKU.