wesight

Why we included this project

Anyone who runs terminal-native coding agents eventually gets tired of juggling separate config files for model credentials, permissions, and per-run metrics. WeSight brings those together in a desktop console that can install or reuse Claude Code, Codex, OpenCode, Qwen Code, OpenClaw, and a few other local CLIs, then gives them a visual chat interface with tool panels, file diffs, and permission prompts. A shared model provider screen saves real time: you configure OpenAI, Anthropic, DeepSeek, Ollama, OpenRouter, or any OpenAI-compatible endpoint once and any agent you launch can use it. Per-run tracking of token usage, time-to-first-token, and tool latency shows which engine actually costs you money. Teams that manage their own local agents rather than relying on one vendor's cloud offering will find it a practical way to keep sessions organized.