1. Running Rollouts on self-hosted machines
We use self-hosted machines for some of our cloud agent workloads and would like Rollouts to run on them as well. On the Rollouts settings page, we don’t see any option to pick self-hosted machines. The only related control we found is the team-wide “Require Self-Hosted Machines” toggle. We can’t turn that on, because some of our users still rely on the regular Cursor-hosted cloud agents, and enabling it would force everyone onto self-hosted.
Is running Rollouts on self-hosted machines supported today?
If it is, how do we set it up without turning on “Require Self-Hosted Machines” for the whole team? For example, is there a per-feature, per-repo or per-user setting?
If it isn’t supported yet, is it on the roadmap?
2. Custom HTTPS MCP servers
We’d also like Rollouts to use our own MCP servers over HTTPS. These are internal tools we host ourselves, not ones from the MCP directory.
Can Rollouts connect to a custom MCP server by URL?
If it can, where do we configure it, and how are auth headers or tokens handled?
Does this also work when Rollouts run on self-hosted machines, for example with MCP servers that are only reachable from our internal network?
In Rollouts settings, select Add data source and pick it.
Set readOnlyHint: true on the tools you want Rollouts to use. It only uses read-only tools.
Only URL-based servers work, not command-based (stdio) ones.
For servers only reachable on your internal network, use the Tailscale integration. It’s in beta on Teams and Enterprise, so ask sales or your account team to enable it.
@Ofir_Cohen, there’s no team rules file like BUGBOT.md for Rollouts today. For now, use Which changes to monitor in settings, and steer a single PR by mentioning the Rollouts handle in a comment (docs).
No timeline on self-hosted support or a rules layer, but we’ll be tracking this thread to gauge interest.