what is shared_url_mcp fr?

This is a small server that does one specific job: it takes a link to a shared conversation from ChatGPT, Claude, Gemini, or DeepSeek and saves the full conversation as a Markdown file. The reason it exists is that standard web fetching tools tend to fail on these links. Each platform has its own obstacle, such as bot protection pages, empty response bodies, or HTML so large and convoluted that the actual conversation text is buried inside it. This tool works around each platform's specific quirk to get the actual conversation content. Installing it involves cloning the repository, entering the folder, and running a single setup command. That command installs what it needs, builds the project, and registers itself with whatever agent programs you already have installed, such as Claude Code, Cursor, or Codex. If a particular agent is not on your computer, it skips it. You then restart your agent program so it notices the new server. For certain links, the tool needs a web browser. It will use your existing Chrome or Edge if available, or it will download a standalone browser on its own the first time it needs one. The server exposes two tools to your agent. The first one fetches the shared link, saves the entire conversation to a file, and gives the agent a short summary of what is in it, including the title, message count, and an outline showing where each message sits by line number. This way the agent can decide whether it needs to read a specific part or the whole file, rather than being forced to process a massive block of text all at once. The second tool does a quick, read-only check on a link to tell you which provider it belongs to, where it redirects, and whether it is still reachable. Output goes into a folder structure with one Markdown file containing the full transcript and a JSON file containing structured metadata. The Markdown file organizes each message as its own section, keeps reasoning inside collapsed blocks, and puts code into standard code blocks.