Most agent frameworks make you wire models, prompts, and execution environments together by hand. Wegent collapses that into one small algebra: Ghost + Shell + Model = Bot, several Bots plus a collaboration mode = Team, and Team + Workspace = a Task you can trace end to end. The unit you reuse is the Bot, not a tangle of glue code.
What Sets It Apart
- A portable Ghost. A Ghost (prompt + MCP + Skills) is decoupled from where it runs, so the same agent intent flows through a Chat shell, ClaudeCode, or Dify without being rewritten. You change execution environments without redefining the agent.
- Declarative definition. Build agents in a web wizard (describe requirements -> AI follow-up questions -> live prompt tuning -> one-click creation) or check them in as YAML. The same team becomes reproducible and reviewable like code.
- Two executor paths. A Cloud Executor runs ClaudeCode and Dify work; a Local Executor connects over WebSocket to host, container, or hybrid setups. Agents can reach private repos and local machines the cloud cannot.
- Wired into real dev flow. It connects to GitHub, GitLab, Gitea, and Gerrit for code, plus DingTalk, Telegram, and other IM tools for human hand-off.
Who It's For And The Trade-offs
Great fit if you want to assemble multi-bot teams over your existing Git hosting and chat tools, and you prefer declaring agents as YAML you can version. Look elsewhere if you only need a single chat assistant -- the Ghost / Shell / Model / Bot / Team / Task layering adds vocabulary you won't use -- or if your execution environment isn't Chat, ClaudeCode, or Dify, since those are the three supported shells.
