Sharing a chat, a model setup or a preset as a link

On tryingopen, anything you share travels inside a link, since there are no accounts to save it to. A share link carries a chat, a setup link carries your choice of models, and a preset saves a setup under a name in your browser.

Share links

Every chat has had a Share button since launch day, August 17, 2026. Press it and the page compresses the conversation and writes it into the link after #c=. The part of an address after the # is called the fragment. Browsers keep it on the device when they load a page, so it stays in your browser and in the browser of whoever opens the link.

That is why we keep no share database. Our privacy policy puts it this way: "There is no copy with us, and anyone you give the link to can read it."

Anyone who has the link can read the chat. Compression makes the link shorter, and any browser that opens it can decode it. Treat a share link like the chat itself: anyone can forward it, and the app you paste it into keeps it like any other message you send.

What the link holds

The page writes the text of each message into the link, and each reply stays credited to the model that wrote it. On the default pick, a single chat can mix replies from the small model on our edge device and from its cloud fallback, and the link keeps each reply's name.

Attachments and reasoning are left out, and so are the details under each reply, such as its stats (what each reply shows covers those). That keeps links short, and it keeps any photo or PDF you attached out of the link.

Sending a share link

On a phone, the button opens your system share sheet, so the link can go straight into a messaging app. On a desktop it copies the link, and the button reads "Link copied".

Long chats make long links. Past 32,000 characters the button reads "Too long. Export instead", because chat apps and mail clients tend to break addresses that long.

Opening a share link

When someone opens your link, their browser decodes the chat and loads it into the page. The page then clears the fragment from the address bar, so a reload or a copied address opens the same setup without the chat. A note says what happened: "Someone shared this chat with you. It arrived inside the link itself, nothing was uploaded."

The reader can reply to pick the chat up where it stopped, or switch models to see how another one answers. A reply is an ordinary turn: the conversation goes to the model along with the new message, under the rules in what tryingopen keeps. A link that got cut short or edited by hand opens a fresh chat.

Export, for a copy you keep

Use Export when a link would be too long, or when you want a file of your own. The Export button ("Download this conversation as a Markdown file") saves tryingopen-chat.md. As the privacy policy puts it, exports "are put together in your browser and saved straight to your device."

The file is plain Markdown, with a heading for each message:

# Tryingopen chat

## You

What is the fragment in a web address?

## GLM 5.3 Flash

The fragment is the part after the # ...

Any picture or PDF you attached appears as a line with its file name.

Setup links

Since October 13, 2026, the address bar describes the models on screen. A provider you pick goes in the path, as in /coreweave, and the model and the small model's cloud fallback go in the query, as ?model=…&cloud=…, each written only where it differs from that page's default. A visit on the defaults stays at a clean tryingopen.com/.

Two examples:

• tryingopen.com/?model=google/gemma-4-31b-it opens on Gemma 4 31B.
• tryingopen.com/?cloud=deepseek/deepseek-v4.1-flash keeps the default small model and sends anything more than a quick question to DeepSeek V4.1 Flash. How d1 routes explains which questions those are.

Copy the address and you have copied the setup. Opening it applies the setup, and the page ignores any part it can't run, such as a model it doesn't offer. If the provider in the path doesn't serve the chosen model, that model's default provider takes over.

Unlike the fragment, the path and query are part of the request a browser sends when it opens a page, so they reach our host, and any app that fetches the link to draw a preview sees them too. They hold only a provider and model IDs; the chat itself stays in the fragment.

Share links carry the setup as well. A share link is a setup link with the chat added after the #, so whoever opens your chat carries on with the same models and provider.

A setup that uses a model on your own computer in Self Chat gets no link, since nobody else could open it.

Presets

Presets shipped the same day as setup links, October 13, 2026. A bookmark button in the composer opens a menu that begins: "Presets are model setups saved in this browser. Each one has a link that opens it."

At the bottom, the menu names the setup on screen ("This setup: …") above a box that reads "Name it (optional)", a Save button and a button that copies the setup's link. Leave the box empty and the preset is named after its parts, for example "LFM2.5-VL 3B + GLM 5.3 Flash on CoreWeave". That one is the small model on our edge device, with GLM 5.3 Flash as its cloud fallback and CoreWeave picked to run the fallback (choosing the company that runs your model covers that last part). Saved presets are listed above, each with buttons to apply it, copy its link or delete it. While one is applied, the bookmark icon fills in and Save becomes Rename.

Presets live in your browser's localStorage, and saving, renaming and deleting all happen there. You can keep up to 20, with names up to 60 characters, and saving a 21st drops the one saved longest ago. So your presets stay in that one browser on that one device, and clearing this site's data removes them.

To share a preset, copy its link. Whoever opens it gets the same models and provider; the name you gave it stays with you.