AISquare
rlx

Tracing (parked)

Session tracing from rlx to AISquare Explainability and the Collective Intelligence test bed are parked since release 0.9.1. What that means, how to check or turn off a setup you already have, and where to go instead.

Parked since 0.9.1

Session tracing through the explainability proxy, and the Collective Intelligence test bed, are parked: unmaintained, out of rlx doctor and out of every screen. Their own commands still run. Do not start a new setup on them. To trace Claude Code to your AISquare workspace, use the proxy integration instead.

Earlier releases could send each fleet agent's model traffic through the AISquare proxy, so every session became a Run in your workspace. Since 0.9.1, rlx keeps that code but shows none of it. Nothing about it is on by default, and nothing you did not turn on starts.

What parked means

  • rlx doctor has no tracing rows, and takes no --target or --project. The same checks run as aisquare explainability doctor.
  • rlx shows nothing of it. A project's tabs are Manager, Board and Settings, with no AISquare tab; the Doctor page has no button that turns tracing on; and nothing in rlx starts a tracing worker. On the phone, an agent's view has Live and Transcript, with no Card tab.
  • The CI test bed stays off unless you turn it on, with AISQUARE_CI or [experiment] enabled = true; the variable wins over the setting either way.
  • aisquare explainability … still works, and rlx launch still traces a session when tracing is on, as before. So do rlx init's --explainability / --no-explainability and the shipping line in rlx status.

If you turned it on before

Tracing you enabled under an earlier release stays on: every agent the fleet starts keeps sending its model traffic through the proxy until you turn it off. Check, then decide:

aisquare explainability status            # is it on, and would the proxy accept a session
aisquare explainability doctor --live     # the checks rlx doctor used to run
aisquare explainability disable           # turn it off for this machine; targets are kept
aisquare explainability reset             # turn it off and remove every leftover ANTHROPIC_BASE_URL of ours
rlx config set explainability.ship false   # stop spooling prompts and board events too

Tracing has two lanes, and disable and reset turn off only the first, the model traffic through the proxy. The second, shipping your prompts and board events ([explainability] ship, which rlx init --explainability turns on), keeps spooling them until you set explainability.ship to false. The spool already written stays where it is.

rlx doctor's anthropic routing row looks for a tracing proxy left in any place an agent reads its base URL from, and names aisquare explainability reset when it finds one.

Tracing also holds back Claude Code's Remote Control: while an agent's ANTHROPIC_BASE_URL points at another host, as the tracing proxy does, rlx launch leaves Remote Control off for it and says why, and rlx doctor's remote control row names aisquare explainability disable as the fix.

The commands that remain

CommandWhat it does
explainability statusThe tracing config, and whether the proxy would accept a session.
explainability doctorThe tracing checks and repairs (--live, --target, --project, --fix).
explainability enable / disableTurn tracing on or off for this machine.
explainability resetTurn tracing off and remove every leftover base URL it set.
explainability use · workspaces · studiosPick where one project's traces land, as the signed-in user.
explainability keyA project's own workspace key.
explainability registerDeclare the machine's agent identities to the workspace.
explainability envPrint shell exports that trace the next run from this terminal.
explainability shipDrain the buffered prompts, notes and task events to the gateway.

Each prints its own --help. They are documented there and no longer here.

On this page