This is an early release preview. You may encounter bugs.
ClawX logo
Unclaimed

Tool interface connectors

ClawX

ClawX is a desktop app that provides a graphical interface for OpenClaw AI agents. It turns CLI-based AI orchestration into a desktop experi

A 87/100 GitHub score ? This grade is derived from GitHub signals, not user votes. Open for the full breakdown.
No votes yet

01 / About

What ClawX is.

ClawX is a desktop application that puts a graphical interface on OpenClaw agents, so configuration, chat, scheduling, and channel management happen in a window rather than at a command line. It embeds the OpenClaw runtime rather than requiring a separate installation, and tracks upstream so new OpenClaw capabilities carry over. Builds are available for macOS 11 and later, Windows 10 and later, and Ubuntu 20.04 and later.

The chat surface keeps multi-session history with streaming Markdown, syntax highlighting, tables, math, direct @agent routing, inline /skill cards, and embedded subagent status with a read-only drill-down into child runs. It previews Markdown, .docx, .pptx, and local HTML files without leaving the app. Agents are created and deleted from the desktop, and deleting a non-default agent permanently removes its managed workspace and chat history.

Around the chat window sit the operational pieces: a visual scheduler for recurring or one-time runs that can insert skills into a scheduled prompt and deliver results to external channels, multi-channel management with per-account agent binding, local-first skill management that discovers skills from several OpenClaw sources and bundles document-processing skills for PDF, XLSX, DOCX, and PPTX, and a provider panel that stores credentials in the system keychain and supports OAuth alongside custom endpoints.

Internally ClawX runs a dual-process architecture. The Electron main process owns the window, system integration, updates, and the lifecycle of the OpenClaw Gateway, while the React renderer talks to a single client abstraction rather than reaching local endpoints directly. Chat itself speaks the Agent Client Protocol over a main-owned stdio bridge, which keeps a stable protocol surface in front of the faster-moving OpenClaw runtime and supports history replay after configuration reloads. Context compaction reserves 25% of a configured model context window, or 50,000 tokens when that metadata is missing.

Features

  • Graphical setup: guided installation and visual settings with real-time validation, no manual config files
  • Embedded runtime: the OpenClaw Gateway ships inside the app and is supervised automatically
  • Chat interface: multi-session history, streaming Markdown, @agent routing, /skill cards, and subagent drill-down
  • Agent management: create, configure, and delete agents, with explicit confirmation on destructive deletes
  • Cron automation: visual scheduling of recurring or one-time runs with skill insertion and channel delivery
  • Multi-channel delivery: independent channels with multiple accounts and per-account agent binding
  • Skill management: local-first installation, discovery from multiple sources, bundled PDF, XLSX, DOCX, and PPTX skills
  • Provider configuration: OpenAI, Anthropic, Z.AI/GLM, and custom endpoints with credentials in the system keychain
  • ACP chat bridge: Agent Client Protocol transport with history replay and recovery across gateway restarts
  • Support export: a ZIP of selected transcripts, sanitized configuration, and diagnostic logs for issue reports

02 / Discussion CREDIBILITY-GATED

Discussion

Reading is open to everyone. Posting and voting need a verified identity or a GitHub grade of B or higher.

  • No discussions yet.

03 / Build

Build with ClawX.

Browse the catalogue for frameworks, tools, and harnesses, each scored on real GitHub credibility.

Get ClawX →

Browse the catalogue