Feature Requests

Please add to the last AI models like GPT 5.5, Opus 4.7, Grok 4.2.

Please add this model it has a 36% hallucination rate which is way lower than GPT 5.5

I have adored Merlin so much and use it very consistently.. I was sad to see that the token limit must have decreased sometime in the middle of the month. I was up to 80% with my usage, which i cannot express how rare that is for me. It usually only goes up to 25%, maybe 50%. It's been really nice to track my usage, as I'm sure many can agree, especially with plan changes. I would have liked a forewarning, maybe a pop up? I want to resubscribe to Merlin because I have LOVED supporting it and having fun with it.. It just makes me very sad to see that the token limit must have significantly dropped without warning. I pay $15 every month consistently for probably half a year and planned to continue to do that, so it was hard for me to cancel the subscription. If any devs see this, please know I appreciate the hard work! It would be beneficial to have some clarity and honesty on the situation. Much love!

On Mac, Cmd+M is still a Windows Minimize Function, it always has been. Cmd+M should be a NO-WAY hotkey. It was reported long ago, by the way. Solution: It should be Ctrl-M on both Windows and Mac. The actual problem: It's still Ctrl+M in my browser, I double-checked it in the browser shortcuts. So when I hit Ctrl+M, I see the new sidebar. (It looks good by the way, just useless, for now.) It says I need to Login (though "Login with Merlin" does not mean "Login TO Merlin", it means what "Login with Google" means, you know; a Google authentication instead of entering email+password credentials. I click on that "Login with Merlin" button and it brings me to the https://www.getmerlin.in/hu/extension-installed web page but there's no login option anywhere. (Yeah, because I am logged in.) Okay, it says Cmd+M (though it's been already changed to Ctrl+M, long ago), I hit Cmd+M, for sure. Nope, it minimizes the browser's window, nothing else happens. Okay, I'm searching for a workaround, visiting getmerlin.in . WHERE I AM LOGGED IN, so the plugin fails to detect I'm already logged in, asking me to log in, instead of chatting with me. It's a big mess now. Btw. I can't wait to see what new system you guys have put together. :)

Perplexity published their model context protocol and implemented the mcp client in their desktop app, now anyone can create their own mcp servers to connect to other applications. Please consider creating a desktop app and adding an mcp client to it.

As everyone is aware, GPT Image 2.0 has already been released. Why, then, are you still providing the older GPT Image 1.5 instead?

Other Multi-LLM offering platforms include even Nano Banana 2 & Pro and other image models as well as video models, Merlin doesnt yet. It feels very outdated and Merlin should up this too. Right now i am dependant on also Google subscription for Nano Banana 2 & Pro. I hope we see some update here too, alot services offer "Unlimited" on their highest plan like 20$/mo or 25$/mo Merlin costs same but we have bit tiny lacking in terms of updating with similar concept multi LLM models.

Many models have disappeared in recent weeks. I really miss Gemini 3.0 and GPT 4 in particular. Is there a good reason for this change? As a paying customer, I would like to be able to choose my models better.

I am reconsidering due to my lack of needs with this service, I intend to wrap up my paid subscription by the end of the following month as I can't justify the incestuous nature of the billionaire class investors using data centers built in areas without proper ecological surveys and the emergence of mass surveillance under a regime that will give you what you want so long as you pay the fee. I'm bummed, but there are not enough safe guards to do this long term and I cannot put money into something people will use to abuse others in the long run. I hope that this changes for the industry and proper rails and guidelines are used, but considering AI was involved in the bombing of Iran, and anthropic being pulled back from it to be replaced with a microsoft AI and then we saw the girl's school bombed right after I feel that people are not being responsible with this learning tool either.

To be able to delete one or more previous requests and their responses and to resume the conversation at an earlier point to modify the angle of the project or discussion, for example. It could also be nice to revisit a part of the discussion (the first 10 exchanges, for example) and test another approach in a new conversation. We could evaluate two different approaches based on the same foundation.