OSS ChatGPT WebUI v4 – Projects, Agent Profiles, Server Tools, 1-Click Sharing

**OSS ChatGPT WebUI** reached **v4** with four headline capabilities: **Projects**, **Agent Profiles**, **Server Tools**, and **1-Click Sharing**. The release is listed under HN AI Agents, points to **https://llmspy.org**, and the related Hacker News thread (**item id 48992035**) sat at **2 points** and **0 comments** when captured. That is a thin early signal, not a popularity metric.

**Projects** suggest scoped workspaces so chats, files, and goals stay grouped instead of living as one long transcript. **Agent Profiles** point at reusable agent configs—system prompt, tools, and behavior—that you pick per task rather than retyping setup each time. **Server Tools** imply server-side actions the UI can call (files, APIs, or host-local capabilities) rather than a pure browser chat box. **1-Click Sharing** is the distribution path: publish a thread, project, or agent setup without a custom export pipeline.

For builders, an open-source **ChatGPT-style WebUI** with projects and profiles is a control surface: you own the host, can pin agent definitions, and can wire tools to your stack without a closed product’s workspace model. Server-backed tools matter when the work is not only generation but side effects—reading repos, calling internal services, or running host jobs under your auth. One-click share shortens the loop from “works on my box” to “teammate can open the same setup.”

Competitively this sits in the open **ChatGPT WebUI / local LLM front-end** lane: ChatGPT-like UX, self-hosted or self-controlled, plus packaging for multi-agent and multi-project use. Closed chat products already ship projects and sharing; OSS UIs win when teams need custom models, private tools, or no vendor lock on history and agents. Early HN traction here is weak (**2** / **0**), so the product claim is clearer than market validation.

Practical next checks on **llmspy.org**: how projects isolate data and auth; whether agent profiles are portable (export/import); which server tools ship and how they are sandboxed; and what 1-click share actually exposes (link, snapshot, live session). Watch whether the HN thread gains substance—real install notes, security review of server tools, and whether sharing is read-only vs editable—before treating v4 as production default.