Two things mattered most to us. First, we could define our own agents with their own instructions, tool access and model choice. We have internal coding standards and a strict security review, so being able to build those into agent roles ourselves, without waiting on the vendor was a requirement. Second, the agents pull context from more than the repo. Most of what we know about our older systems lives in wiki pages and old design documents, and the agents actually use it. The maintainability tracking also gave us a baseline we never had for these codebases. Review collected by and hosted on G2.com.

Actually I'd like a wider choice of models. We have an approved list of providers internally, and not all of them are supported by Revolte yet, which limits a few teams. The Revolte team has been open about their roadmap, so I expect this to improve. Review collected by and hosted on G2.com.