Design goals
ReplayCore adds negligible overhead. The recording pipeline runs on dedicated background threads, off the main server thread, so it does not compete with tick processing for CPU time. Per-server RAM overhead is roughly 30-80 MB depending on player count and upload queue settings.
If you observe TPS impact, it is almost always I/O or configuration related rather than the capture itself. The steps below target the common causes.
Reduce upload contention
If uploads compete with game traffic on a constrained connection, use Configuration Studio's Connection section to cap upload throughput, leaving headroom for players. Reducing capture quality in the managed Recording policy also reduces the data captured and uploaded.
Bound archive sizes
On long-running servers, configure rotation in Configuration Studio's Storage section so each match or threshold window seals into its own archive. Smaller archives upload sooner and keep local disk usage bounded. If uploads stall entirely, confirm outbound HTTPS to api.replaycore.com on port 443 is not blocked.
Still seeing impact?
Capture a timings or spark profile during a busy period and contact admin@forgevector.co.uk with the report, your Minecraft version and platform, the ReplayCore version, and the affected server name. Never send the recording key or private plugin state. The report lets the team pinpoint whether the cost is in capture, compression or upload.