Skip to main content

​
Global stream

muxAI streams agent activity in real time via Server-Sent Events (SSE). A collapsible Stream panel sits at the bottom of every page, showing live output from all running agents. The stream shows:
  • Run start/end events with agent name and status
  • Log lines as agents think and call tools
  • Tool calls formatted as tool_name(args)
  • Errors when something goes wrong

​
Stream History

The Stream History page (/streams) gives you a full-page view of both live activity and past runs. It combines:
  • Live entries from the current session (stored in browser localStorage, capped at 500)
  • Database history loaded from past runs below a divider
This is where you monitor what’s happening across your entire agent fleet.

​
Which agents appear in the stream?

Only agents with persistLogs enabled show up in the stream and history. This is controlled per agent in the edit form. By default: This keeps the stream focused on team-level outputs. Reporter activity is still captured in each reporter’s individual run history. You can always view a specific reporter’s runs from their detail page.
If you want to debug a specific reporter, temporarily enable persistLogs on that agent. Remember to turn it off when done to avoid cluttering the stream.

​
Run history

Every agent run is recorded in the database as a HeartbeatRun with:
  • Status (succeeded, failed, cancelled, timed out)
  • Full log output
  • Result JSON (if the agent produced structured output)
  • Outcome label and custom fields (user-marked after the fact — e.g. Win / Loss / NA plus pnl, note)
  • Timestamps and exit code
  • Session ID (for resume support)
  • Optional parentRunId linking the run to a prior decision being re-examined (see Memory & Decisions)
View an agent’s runs from its detail page or the Runs tab. Each run shows the complete log and, if available, a formatted result card. You can mark a run’s outcome from its detail page; trading agents with reviewDecisions will see these outcomes on their next run via get_my_decisions. Parent runs that have been re-examined show a Conviction history list of their child re-examinations underneath the result card.

​
Results

The Results page (/results) shows all runs that produced structured JSON output. Each result is rendered as a card based on the agent’s resultCard configuration (e.g. Trade Decision, Research Summary, Alert). For trade decisions specifically, /results/trade-decisions is a dedicated terminal with a blotter, candle chart with entry/TP/SL overlays, and per-trade controls. Re-examination runs don’t appear as standalone rows — their latest conviction score and suggested action are surfaced inline on the parent trade row, and the Re-examine button on each active trade lets you trigger a fresh full-council re-evaluation without leaving the page. The Dashboard also shows the latest result from your most recent run.