Live proxy and collect
A separate advanced workflow, apart from the offline check and the model comparison: point an endpoint at RedCrown, let it forward the calls and capture a sample, then replay the captures through an evaluation when you ask for it. It needs a Growth or Scale plan. Endpoints are created through the API or MCP and managed in the app. Nothing here runs on a schedule: a replay happens when you start one.
1. Set up the proxy
Create a proxied endpoint. RedCrown forwards each call to your incumbent model and stores a sampled fraction of the calls as captures, encrypted, for a later replay. This is the one path where RedCrown sits in the request path of your application; every other path on these docs reads files or calls providers on your behalf.
Endpoint creation is not available in the UI at any plan tier; it is API and MCP only. Once an endpoint exists, open Advanced and navigate to Live traffic (/advanced/live-traffic) to manage it: start a replay, route a percentage of calls to a chosen candidate, or open the reviewer panel. The proxy is a Growth-and-up capability; Starter and Free show an upgrade prompt instead of the endpoint list.
Not available via the CLI. Use MCP or call the API directly.
Call setup_proxy with your target URL:
{
"target_url": "https://your-model-endpoint.example.com/v1/transcribe",
"experiment_id": "<experiment-id>"
}The response includes the endpoint key and the forward URL. Update your application to send traffic to the forward URL instead of the upstream directly. experiment_id, sampling_rate, and daily_budget_usd are all optional.
2. List endpoints
Review all proxied endpoints in your workspace, including their capture counts and sampling rates.
The Live traffic page (/advanced/live-traffic) lists your endpoints by key and routing mode. Select one to see its sampling rate, daily budget cap, and canary routing percentage.
Not available via the CLI. Use MCP or call the API directly.
Call list_proxies with no arguments to retrieve all endpoints in the current workspace:
{}Each entry includes id, endpoint_key, target_url, sampling_rate, daily_budget_usd, status, and active_route.
3. Replay captures
Fan captured traffic across every candidate model and store the scored outputs for comparison or reviewer sign-off.
On the Live traffic page, select an endpoint and click Run shadow eval. RedCrown fans the captured inputs across all candidate models and links the results to the endpoint.
Not available via the CLI. Use the app or MCP.
Call replay_captures with the endpoint ID:
{
"endpoint_id": "<endpoint-id>"
}RedCrown decrypts each stored capture, fans it across the non-incumbent candidates, scores relative to the incumbent, and persists one result row per capture per candidate.
Last verified .