Privacy
Self Chat saves your chats only where you choose
Self Chat keeps each chat in your Google Drive or a folder on your computer, and sends each cloud message under a zero-retention requirement.
Tryingopen's site description promises "No account, nothing saved." Keeping a chat from the ordinary page means a share link or an export. Self Chat, at tryingopen.com/self-chat, is the mode for keeping your chats in a place you control. Its title is "Self Chat: your chats, zero retention", every page's header links to it, and the older /zdr address redirects there.
Where a chat goes
The page states the promise plainly: "Every chat gets its own folder in Open Chats, in the place you pick." Each chat folder is named after your first message and holds up to four things:
Open Chats/
Plan a weekend in Lisbon/
Plan a weekend in Lisbon.json
Plan a weekend in Lisbon.md
Plan a weekend in Lisbon.csv
artifacts/The JSON is the record the page reads back. The Markdown and the CSV (one row per message) are readable copies, rewritten from it after every finished reply. The artifacts folder holds the images you attach and the code and files the model makes.
The page adds: "Nowhere else: not in this browser, not on our servers." Your browser talks to the storage service directly, so the chat files never pass through tryingopen.
The two places that work today
As of October 2026, the page offers two:
| Place | What the site can see | How you connect |
|---|---|---|
| Google Drive | "This site can see only the files it made in your Drive." | Press "Connect Google Drive" and sign in |
| A folder on this computer | "This site sees only that folder; nothing is uploaded." | Pick a folder you have, or make a new one |
The chat opens once you connect one of them. Until then, the chat list reads "Your chats show up here once they have somewhere to live."
Google Drive
Self Chat asks Google only for the drive.file permission. The privacy policy explains what that allows: Self Chat can "see and change only the files it created," and uses that "solely to create, read, update and move to the trash your chat files in the 'Open Chats' folder." The access token stays in the browser tab you are using. Deleting a chat from Self Chat moves it to Drive's trash.
To take the permission back entirely, use your Google account's third-party connections.
A folder on your computer
This option uses a browser feature that lets a page read and write one folder you pick. Chrome, Edge and other Chromium browsers on a computer have it. Elsewhere, the card says "Needs Chrome, Edge or another Chromium browser on a computer." On a later visit the browser may ask you once more to allow the folder.
The page names each place's limit. A chat in Google Drive is a file Google holds like any other file there, under Google's own privacy policy. For the folder, it says: "A folder on this computer is held by no one but you. Delete the folder and the chat is gone."
Coming soon
The page lists OneDrive, Dropbox and Box under "Coming soon", with no date given.
Stricter rules for every message
Self Chat applies tryingopen's strictest privacy rules, and our server enforces them on each request.
Zero retention for every message. In the page's words: "Each message goes straight to OpenRouter with a zero-data-retention requirement attached. Only a provider that promises to delete it after answering can serve it, and if none will, the message fails instead of being sent anyway." Self Chat never uses the ngrok gateway that some models take by default on the ordinary page. What zero data retention covers explains the requirement and its edges.
Fewer models. The picker keeps only models OpenRouter can serve under that requirement. As of October 2026 that is 15 cloud models, and the picker's header says so: "15 models here, each with a provider that promises zero retention. Anything without one is left out."
| In Self Chat | Left out |
|---|---|
| GLM 5.3 Flash, GLM 5.3, Qwen3.8 27B, Qwen3.8 2.4T, Nemotron 3.5 Lightning, Nemotron 3 Ultra, DeepSeek V4 Flash, DeepSeek V4 Pro, DeepSeek V4.1 Flash, Muse Glimmer 30B, Inkling, Inkling Small, Hy4 Preview, Ling 3.0 Flash VL, LFM2.5 24B | Both edge-device picks (LFM2.5-VL 3B and Gemma 4 E2B), Gemma 4 31B, Gemma 4 26B, GPT-OSS 120B, Qwen3.8 Flash, Mistral Large 4, MiMo V2.6 Flash, MiMo V2.6 Pro |
The edge-device picks reach our own hardware through a tunnel. Gemma 4 31B, Gemma 4 26B and GPT-OSS 120B are set up around a second route, to Amazon Bedrock through ngrok's gateway, so Self Chat leaves them out, even though a chat opened directly on one of them on the ordinary page starts on CoreWeave. Qwen3.8 Flash, Mistral Large 4 and the two MiMo models have no route on OpenRouter that meets the requirement.
Five models labelled "Provider may log" on the ordinary page do qualify (GLM 5.3 Flash, GLM 5.3, DeepSeek V4 Pro, DeepSeek V4.1 Flash and Hy4 Preview), because here they always go through OpenRouter under the requirement. You can still pick the company that runs a model: /self-chat/coreweave opens with CoreWeave chosen.
No PDFs. A PDF would be read by OpenRouter's file parser, which sits outside the zero-retention requirement, so Self Chat refuses PDFs. Images are fine: "Images go straight to the model, inside the same zero-retention request," and a copy lands in the chat's artifacts folder.
Search is your call. Web search is off until you switch it on in the composer. The switch spells out the trade: "Turn it on to let the model look things up; searches go to a search provider, outside zero retention." How web search works has the rest.
No side-by-side. On the ordinary page, picking a cloud model also sends your first two questions to our small model for comparison. That never happens in Self Chat (how the side-by-side works).
Every reply made under these rules carries a "Zero retention" badge, so a saved chat still shows which answers met them.
A model on your own computer
Self Chat can also use a model running in Ollama, LM Studio, llama.cpp or Jan on your own computer. From the page: "Pick a model on your own computer and this page talks to it directly. Those messages never touch the internet at all."
The button reads "Use a model on this computer", and those replies are marked "On your computer". If you switch search on with a local model, your latest message goes to our server for one lookup, and in the privacy policy's words, "only the results go back to your computer."
Three of the four apps need one setting changed before the page can reach them; llama.cpp's server works as it is. Use Ollama, LM Studio, llama.cpp or Jan from Self Chat walks through each one.
Who it's for
Pair a local model with a local folder, leave search off, and the whole chat stays on your machine. Pair a cloud model with Google Drive and each chat sits in your Drive, ready to reopen in Self Chat or read as Markdown. What tryingopen keeps covers the rest of the site.