It proposes work
A reporter may request sources, a Scheduler may draft an appointment, or an engineer may request a limited command. The role determines which workflow and data it should see.
Clawsolator is a XERXES SI product concept for running specialized AI agents under human-defined operating boundaries. It organizes the agent experience while XMAN provides a separately developed manager and approval workflow. The goal is not a smarter prompt. It is a safer relationship between an agent request, a permission, and an executable action.
Clawsolator is the product-facing experience for governed agent teams. XMAN (XERXES Agent Manager) is the local management component that coordinates requests and approved actions around an OpenClaw/rootless-Podman environment. These are complementary layers, not replacements for upstream security patches or a guarantee of perfect isolation.
A reporter may request sources, a Scheduler may draft an appointment, or an engineer may request a limited command. The role determines which workflow and data it should see.
Approved tool access, container boundaries, filesystem access, networking and resource envelopes must be implemented as real controls, not text in an agent personality file.
The XMAN development source includes a limited daemon-mediated request path, resource policy and special-request review. Host qualifications are required before describing a particular installation as secure.
Gateway Worlds, domains, offices, specialist roles and visible approval states help an operator understand what is running and why an action was allowed.
No. OpenClaw is an upstream agent platform with gateways, tools, channels, skills, security settings and its own isolation options. Clawsolator is an independent XERXES SI product concept that seeks to make governed deployment and team operation easier to inspect. OpenClaw's security features and current patches are still required.
Clawsolator describes the operator-facing product and agent office system. XMAN is the manager implementation whose authority checks, requests and resource policies can be inspected in the source archive. An attractive dashboard is not proof of enforcement: the manager, gateway and rootless runtime must each be configured and tested.
A visual world represents a site or deployment. A gateway instance has its own state and configuration; an office is a way to organize specialized agents and permissions. Separate gateways can be useful when trust domains must not share a gateway configuration. A different room graphic by itself does not create a security boundary. Explore the Gateway Worlds architecture.
The XMAN 0.4.76a1 source archive documents ./xman.sh autosetup, approval handling and bounded management interfaces. That establishes source-level implementation, not an independent production certification, external penetration-test result or guaranteed deployment success. The website's animated shell is a redacted demonstration, never a claim of live execution.
Four concise paths through the architecture—written for operators, developers, security teams and evaluators. Each distinguishes configured controls from goals that still need testing.
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.