THE CLAWSOLATOR PRINCIPLEA prompt is not a permission system.Meet the independent daemonOperator authority belongs outside the agent.
THE FIVE-STEP OPERATING EXPERIENCE

From fresh machine to
governed agents.

Clawsolator is designed to combine proven upstream technology into a deliberate, owner-controlled setup sequence: inspect, install, connect, isolate, and verify.

05Guided stages
instead of guesswork

Watch the redacted XMAN autosetup demonstration →

GUIDED SETUP

Make secure deployment feel obvious.

No agent should need host root access just to finish onboarding. Where privileged host changes are needed, the operator—not the model—must decide.

Proposed product presentation shown here: actual supported actions depend on the installed OpenClaw and host versions.

01

Inspect before touching anything.

Detect OpenClaw version, existing Gateway, active services, identity boundaries, hardware and supported Podman capabilities.

02

Install or repair carefully.

Preview changes, protect existing state, avoid duplicate tunnels or overwriting foreign resources.

03

Connect through protected routes.

Configure Cloudflare Access for protected browser access; use Tailscale for private paths where applicable. Verify TLS and authentication separately.

04

Isolate individual agents.

Provision or select session-scoped rootless Podman environments, bound workspaces and narrow tool privileges.

05

Prove the outcome.

Read back actual settings, container identity, isolation and network path; store diagnoses and provide recovery.

WORKING MODEL / SIMULATED CONSOLE

Let the agent ask. Let the daemon decide.

An interactive illustration of the XMAN policy path. This browser demo does not run Podman or connect to the actual daemon.

XMAN / ROOTLESS POLICY CONTROL SIMULATED
AGENT-001 · CAPABILITY REQUEST
memory += 32 MiBReason: task needs additional headroom
POLICY & EXECUTION RECEIPT
request.received
identity verified
waiting for evaluation_
READY FOR TEST
AUTHORIZATION IS NOT EXECUTION

One rootless daemon. One deliberate escalation boundary.

Automatic CPU/RAM policy is not the same thing as granting root. The planned privileged executor would remain a separate, small, explicitly authorized host service.

DEMONSTRATED RESOURCE PATH

Worker → XMAN → Podman

Request preparation, commit, daemon policy, live resource adjustment, Podman readback and durable outcome reporting.

DEMONSTRATED IN A LIMITED LIVE TEST
FUTURE PRIVILEGED PATH

Operator → narrow executor

Immutable script staging, independent human authentication, one-use signed authorization, host-bound execution and crash reconciliation.

ARCHITECTURE / NOT YET SHIPPED
FUTURE DEFENSE

Security and interruption daemons

Additional scoped monitors could identify runaway work, enforce pause/containment controls, and reconcile interrupted operations.

ROADMAP / NOT VERIFIED
THE NEXT STEP

Designed for agents that can do more.

The security system must become more disciplined as autonomy increases.

SECURE OPENCLAW INSTALLATION

From installed to trustworthy: an operator checklist.

BEFORE INSTALL

Scope the environment

Identify the host, users, network reachability, data sensitivity and which actions agents are actually permitted to perform.

BEFORE REMOTE ACCESS

Verify identity & exposure

Use reviewed Cloudflare Access or Tailscale patterns, confirm TLS and pairing, and keep the gateway off the unrestricted public internet.

BEFORE AUTONOMY

Test the boundary

Verify rootless runtime constraints, deny a disallowed operation, review a sensitive request and confirm an allowed resource change by readback.

Watch the privacy-screened autosetup →

Straight answers / technical briefing

Questions engineers and executives actually ask

How does a governed agent action proceed?

The agent proposes a scoped request. XMAN compares it with an approved policy and resource budget. Routine compliant actions can proceed; exceptional requests require operator review or are rejected. The manager then verifies the resulting runtime state and records an audit trail.

Why use rootless Podman?

Rootless Podman gives containers user-scoped execution rather than ordinary host-root power. Resource restrictions, network boundaries, writable mounts and credentials still require careful configuration. Rootless alone is not a complete security system.

Where do Cloudflare and Tailscale fit?

They can provide protected access and private network paths for supported deployments. They do not replace agent sandboxing, least-privilege tool permissions, approval workflows or auditing.