A Visual Workspace for CLI Agents That Refuse to Grow a GUI

What it does

WeSight is an open-source Electron desktop console for local AI coding agents. It detects or installs terminal CLIs such as Claude Code, Codex, OpenClaw, and DeepSeek-TUI, then surfaces them through a unified graphical workspace with chat, file diffs, tool panels, and permission prompts. Model providers—OpenAI, Anthropic, Ollama, OpenRouter, or custom endpoints—are configured in one place and mapped into whichever engine is active.

The interesting bit

Rather than inventing another agent, WeSight treats existing terminal-native CLIs as backends and wraps them in a React frontend. The result is a live metrics dashboard tracking TTFT, TPS, token burn, and tool latency, plus IM routing and a desktop pet, all layered on top of shell-bound tools you already use. It is largely glue code, but it is ambitious glue: a management layer that adds a GUI, scheduling, memory, and a skill marketplace to programs that normally live in a terminal.

Key highlights

• One-click setup or auto-detection for eight agent engines, including Claude Code, Codex, and Hermes Agent.
• Centralized model provider routing across official APIs and OpenAI-compatible endpoints.
• Live workspace with file writes, code diffs, generated artifacts, and a runtime metrics dashboard.
• IM Agent Hub for routing Feishu messages into selected engines with per-bot profiles.
• SkillHub marketplace, scheduled tasks, session memory, and a desktop companion widget.

Caveats

• Early releases are macOS Apple Silicon only; other platforms are not yet available.
• Preview builds are not signed or notarized, so macOS Gatekeeper quarantines the app and warns that it is “damaged.”

Verdict

Worth a look if you juggle multiple local coding agents and want a single desktop hub for model routing, cost tracking, and chat. Not for you if you are not on a Mac with Apple Silicon or if you prefer staying in the terminal.