CLAWSOLATOR ALPHABeta release planned · License under considerationRead what is verified
XERXES SI / SECURITY ENGINEERING

AI Agent Security Comparisons: OpenClaw, OpenShell, Hermes and More

A serious comparison names the protections competitors already ship, identifies the questions they do not claim to solve, and shows what XMAN adds at the operator layer.

Independent engineering commentary by XERXES SI · Updated November 1, 2026 · Claims separated from roadmap

Fair technical comparisons

What each product solves—and what remains to be evaluated

The following are not allegations of product flaws. They explain the published security scope, operational fit and where independent governance may add value.

TECHNICAL SCOPE COMPARISON

OpenClaw

What it does: Provides agent tools, gateways, native Podman sandboxing, approval semantics, documentation and an active patch program.

What to check: Its explicitly trusted-operator gateway model may not align with a multi-tenant business environment without additional isolation and operations design.

Clawsolator’s role: Use XMAN as an operator layer for controlled host requests, durable approvals, policy-driven resource planning, gateway oversight and recoverable jobs.

OpenClaw security status & trust model ↗

Explore this boundary →

TECHNICAL SCOPE COMPARISON

NVIDIA OpenShell / NemoClaw

What it does: Provides an externally enforced secure runtime, policy-managed filesystem/process/egress, credentials, audit logs, and approvals.

What to check: It is not positioned as a substitute for every organization-specific business workflow, master-agent assignment, operations menu, or resource-budget process. This is a scope distinction, not a vulnerability.

Clawsolator’s role: Potential complementary deployment: OpenShell enforces runtime policy while XMAN coordinates people, resource and change-management processes; integration is planned, not shipped.

NVIDIA OpenShell product and policy architecture ↗

Explore this boundary →

TECHNICAL SCOPE COMPARISON

Hermes Agent

What it does: Documents user authorization, approvals, write safety, container isolation, credential filtering, scanning, session boundaries and sanitized tools.

What to check: Its documented safeguards are meaningful; compare how its runtime approval and session model maps to a separate organization-level operations approval ledger.

Clawsolator’s role: XMAN can add a separately scoped operator workflow to managed host actions when integrated; compatibility and parity are unverified.

Hermes Agent security guide ↗

Explore this boundary →

TECHNICAL SCOPE COMPARISON

ZeroClaw

What it does: Reports supervised autonomy, command policies, OS sandboxes and cryptographic receipts in its own architecture.

What to check: Security design still must be evaluated per installation; optional escape hatches change the risk posture.

Clawsolator’s role: Compare role enrollment, durable operator approval workflows, audit exports and resource planning to the organization’s requirements.

ZeroClaw published architecture ↗

Explore this boundary →

TECHNICAL SCOPE COMPARISON

Docker / Podman / Kubernetes

What it does: Mature isolation and orchestration foundations with extensive hardening options.

What to check: A container alone does not define business approvals, agent intent policies, secrets brokering, independent spend limits, or incident response.

Clawsolator’s role: Use rootless containers as substrate and XMAN for the operator-level requests, budgets and approval lifecycle.

OpenClaw rootless Podman backend ↗

Explore this boundary →

TECHNICAL SCOPE COMPARISON

Managed AI workflow platforms

What it does: May offer valuable managed integrations, support, procurement review, compliance documents, and rapid deployment.

What to check: Customers must verify visibility, data retention, contractual rights, log export, network policy, service boundaries and portability instead of assuming security either way.

Clawsolator’s role: Offer inspectable self-managed operations as an alternative where custody and reproducibility matter.

Gumloop Trust Center ↗

Explore this boundary →

The enterprise procurement question

Can you show us what the agent could reach, who authorized an exceptional operation, what actually ran, and what the system recorded when it failed?

Security is a combination of identity, runtime enforcement, configuration, human workflow and evidence—not a ranking badge. A good comparison runs the same scenario in each candidate environment, records results and documents missing coverage.

Use the engineering verification checklist →

Strong claims deserve strong evidence.

Inspect the controls, verify the boundaries, and ask what happens after a failure—not merely before it.

Explore the security lattice →Read the architecture →

Compare the right categories before you buy

OpenClaw, OpenShell, Hermes, ZeroClaw, Gumloop, Aider, Continue and cloud providers solve overlapping but different problems. Find what each actually provides, what you must still configure, and what XMAN adds.

NEW RESEARCH / DOCUMENTED OPERATOR FRUSTRATIONS

What users asked the systems to do

Firsthand issue reports, exact failure symptoms, the proposed engineering response and explicit verification limits.

Engineering humor with sources →