clawwork-ai/ClawWork

What it solves

ClawWork provides a dedicated, local-first workspace for AI agents, moving beyond the limitations of simple chat interfaces. It solves the problem of "chat-history mud" where tasks, files, and context are lost in long threads, by treating every agent task as a durable, isolated workspace with its own session, artifacts, and history.

How it works

ClawWork acts as a client for the OpenClaw runtime. It connects to an OpenClaw Gateway via WebSocket to run agent tasks in parallel. The application uses a three-panel UI (task list, active work, and artifacts/context) to organize work. All data—including messages, tool calls, and produced artifacts—is stored locally using SQLite and Git, ensuring 100% local data ownership.

Who it’s for

Users who run multiple AI agents or complex, long-running agentic workflows that require structured tracking of progress, durable artifact storage, and the ability to manage multiple parallel sessions without mixing context.

Highlights

• Task-First Workflow: Run parallel tasks in isolated sessions with support for scheduled (cron) execution.
• Multi-Agent Teams: Package multiple agents into coordinated ensembles (coordinator and worker agents) with roles and personalities.
• Local-First Architecture: All session data and artifacts are stored locally; no cloud sync or external databases are required.
• Enhanced Visibility: Real-time streaming responses, inline tool call visualization, and a dedicated cost/usage dashboard.
• Advanced File Handling: Auto-extraction of code blocks and images from assistant replies, and the ability to watch up to 10 local folders for automatic context indexing.
• Cross-Platform: Available as a desktop app (macOS, Windows, Linux) and a Progressive Web App (PWA).