---
schema_version: "newruntime-agent-readable-v0.1"
type: "system_dossier"
slug: "openclaw"
title: "OpenClaw"
description: "OpenClaw is tracked as a living experiment in always-on operational agents for channels, tools, previews, approvals, and bounded publication workflows."
status: "published"
system_status: "running"
one_line_role: "Always-on operational agent surface for channels, tools, and persistent workflows."
research_question: "How close are always-on agents to reliable channel-native operation with bounded approvals and visible traces?"
current_observation: "Useful for owner-gated publication handoff and private previews, but public write actions must remain explicitly approved."
verification_level: "workflow-tested"
last_verified: "2026-07-21"
related_shifts: ["goal-scoped-agent-loops"]
related_posts: ["claude-code-long-running-loops","claude-hud-agent-observability"]
topics: ["agent-systems","openclaw"]
source_urls: []
public_boundary: "Sanitized dossier. Public routes intentionally exclude private operational commands, credentials, internal handoffs, and private discovery infrastructure."
routes: {"html":"https://newruntime.com/systems/openclaw/","markdown":"https://newruntime.com/systems/openclaw.md","json":"https://newruntime.com/systems/openclaw.json"}
source_format: "markdown"
---

# OpenClaw

This public dossier is intentionally sanitized. It should not expose deployment
commands, secrets, ports, tunnels, private logs, or access policy.

## Current public verdict

OpenClaw is useful as an owner-gated operational surface. It should not be put
in the routine card-processing path unless the task genuinely needs deep work or
publication handoff.