SweepAI is shutting down

[Anthirian]

This email was sent out to the Sweep userbase today:

To our Sweep community,

After careful consideration, we have made the difficult decision to shut down Sweep. We are incredibly grateful to each of you for your support and for being part of this journey — it has truly meant a lot to us.

Here's what you need to know:

• Servers go offline: April 9th
• New subscriptions: Disabled as of today

We decided to shut down because autocomplete usage is declining industry-wide as agents have become significantly better. Competing in the agent space has also been an uphill battle against foundation model labs with far greater resources — such as heavily subsidized plans like Claude Code Max.

That said, we're not walking away without leaving something behind for the community:

1. We'll release a locally running version for both agent and autocomplete.
2. We'll also open-source our latest production 7B autocomplete model and low-latency runtime.

[leonardcser]

Thanks for creating this issue. For those who want to use an API, the mercury api provider is similar in quality (though we could improve the context sent to the endpoint).

[tkataja]

Thanks for creating this issue. For those who want to use an API, the mercury api provider is similar in quality (though we could improve the context sent to the endpoint).

Thank you for this awesome plugin to nvim. For long I thought Cursor Tab was their most valuable thing for me, and just love it to get similar experience on my vim. I was really liking the Sweep API but moved to Mercury API after this announcement of shut down. What is you gut feeling on the priority to enhance the context improvement for Mercury edits?

[leonardcser]

The mercury API has no direct support for LSP diagnostics but I will do some testing if I can still include it in the request. Improvements to the mercury api coming soon too :)

[leonardcser]

A quick heads up, the free plan Mercury has a 100 req/s rate limit, which is easy to hit.

You can increase the text_change_debounce to 200-300ms.