Skip to content

Machines, statistics, and tools

This page lists the Machines, Statistics, and Tools tabs and the Config Flags pane.

Config Flags pane

Open SASE Admin Center with #, then Config > 02 Flags. The pane is the keyboard-first control surface for every code-owned SASE feature flag. It shows effective state, kind, default, provenance, saved machine preference, description, removal bead, and removal horizon; it does not edit portable configuration files.

Config's nested catalog is:

01 All · 02 Flags · 03 Holds · 04 Launch · 05 Memory · 06 Snippets · 07 Macros

sase flag enable and sase flag disable remain available alongside the pane as the automation surface.

The pane uses the Admin Center list/detail layout: a header with registered/on/saved counts, a flag rail, a scrollable detail card, a hidden inline filter, and a one-line footer. Effective on/off and source are shown separately from saved on/off. When environment or root CLI -f/-F still wins, a yellow “forced for this process” warning explains that saving restarts sase's TUI and the service host but will not take effect until that override is removed.

Key Action
/ Open an inline filter over key, description, kind, effective state, and provenance
Esc Close/clear the filter before closing Admin Center
j / k Move the rail; arrows, Home/End, and mouse clicks work the same way
q Close Admin Center
Enter/Space Open a cancel-first confirmation (OFF -> ON or ON -> OFF)
r Reload the catalog
0 then 1–7 Jump to a numbered Config child

Confirmation is cancel-first. It names the flag, the current-to-target state, the saved state path, any shadowing source, and that sase's TUI and service host restart after TUI tasks and installation changes finish. Confirming writes only the machine-state file, waits up to 60 seconds for TUI-local work and installation mutations, then performs one controlled sase's TUI+service-host restart. Independent commands keep running. Durable operations started from this TUI still finish their result handling first, within the same 60-second wait. A restart failure does not roll back the saved preference.

See feature flags for precedence, the state file, and the CLI contract.

Machines Tab

Open the SASE Admin Center with #, then press 3 or select Machines. This is a local inventory and administration surface for the controller plus every enrolled remote. It does not contact all gateways when opened: remote rows begin with state not checked and health unknown until you request an explicit status check. Remote capacity remains not reported even after that check because the status response does not carry runner capacity. The local row reports the controller's configured runner capacity immediately.

The list shows Alias, State, Health, Capacity, Last observed, and Endpoint. The local Alias is the controller's configured machine name even though its Agents query token is here. The detail card adds capabilities, provider, pinned installation identity, quarantine state, and the status message from the most recent check while this Admin Center remains open.

Key Action
j / k Move selection
/ Filter aliases, endpoints, providers, installation identity, and quarantine
s Run one bounded authenticated hello for the selected remote
c Show the persistent connect/enrollment flow
r Show repair guidance for the selected remote
R Show the rename command for the selected remote
x Show removal guidance for the selected remote
y Copy the commands shown in the action card
Enter Close Admin Center and open this machine's agent tab (or filter Agents when the tab strip is hidden)
f Close Admin Center and open Agents filtered to this machine
U Reload local machine inventory without probing gateways

Connect, repair, rename, and remove are guidance actions: they display canonical CLI commands and safety notes but do not mutate machine state themselves. Enter selects the machine's agent tab when the tab strip is visible, and otherwise applies an Agents query of machine:local for the local controller or machine:<alias> for a remote (f always filters). See the Remote Dispatch Runbook for the credentialed enrollment and recovery procedures.

Statistics Tab

Open the SASE Admin Center with #, then press 6 or switch to Statistics. Its eight sub-tabs summarize overview, runners, projects, providers, agent activity, macro usage, plan/question activity, and performance for the selected time range. The strip is numbered 01 Overview · 02 Runners · 03 Projects · 04 Providers · 05 Activity · 06 Macros · 07 Plans & Questions · 08 Perf; press 0 and then the second digit to jump straight to a view. The Admin Center-wide ' entry-jump key arms this same numbered-view selection instead of painting row hints — Statistics has no row cursor, so the already visible strip numbers act as its jump hints; Esc or any non-digit cancels. Use [ / ] to move between views, t / T to cycle time ranges, c to enter a custom range, p / P to cycle project scope, Ctrl+D / Ctrl+U to scroll, r to refresh, and ? for the pane's key help. First open seeds the project filter from the current project when ace.current_project.seed_filters is on; p / P can always cycle away from that seed, including back to All projects. On Overview, Agents Run, Success Rate, and Commits open Projects; Plans Proposed and Questions open Plans & Questions.

The Perf sub-tab combines five headline measures—Startup (median visible-ready), Stalls (stall count, with hitches named separately in the tile's detail line), Launch (p95 total launch time), Agent p95, and LLM p95—with startup stages, stall/hitch events, grouped latency and reliability, and source-coverage diagnostics. Press g to group latency by subsystem, provider, or workflow; the grouping also decides what the count column counts (LLM invocations, agent runs, or an ungrouped count) and whether a Share column applies. Perf counts come from telemetry and TUI logs, not the artifact index, so they are not comparable with the run counts on the other sub-tabs. Perf is global: the project chip remains visible but is marked not applied. See Reading the Admin Center Perf view for data sources, retention, and probe details.

The Statistics Macros sub-tab reports macros referenced by agent launch prompts:

  • By Usage ranks macros by runs and shows references, share, agents, success, runtime, and recency.
  • By Model breaks each macro down by model.
  • By Project breaks it down by project.
  • Used With shows macros referenced together in the same run.

Press g to cycle those four groupings without reloading the underlying statistics. Press x to choose one macro and replace the ranking with its full time, model, project, provider, tribe, and co-usage breakdown; press X to clear that focus. The range and project filters apply before all macro aggregation.

These counts come from each run's launch-boundary macros.json, recorded before prompt expansion. A run counts once per macro name, while Refs counts distinct argument variants of the same name separately. References introduced inside workflow step templates are intentionally excluded. Historical runs appear after the agent-artifact index rebuilds at the current schema.

Each record carries a kind — workflow, part, or swarm. A macro swarm is consumed by the dispatcher before any agent starts, so it never appears as a lexical reference in a child's prompt; instead the swarm is attributed to every agent it launched, and a nested swarm records every link of its chain. Because swarm records carry no arguments, Refs equals Runs for a swarm row. Attribution is forward-only: runs launched before this feature shipped are not backfilled.

This Statistics sub-tab is distinct from Config's Macros child described in Macro Browser: that child browses and edits macro definitions, while the Statistics sub-tab measures how launch prompts used them.

Tools Tab

Open the SASE Admin Center with #, then press 7 or switch to Tools. Its three sub-views cover Runs (silent runs pinned first in red, then live, then settled newest first), Failures (failure-signature groups with run and agent witness counts), and Catalog (the current project's named tools with LAST and TYPICAL from the run ledger). The pane is filtered to the current project by default; press A to widen it to all projects. Use [ / ] to move between views, / to filter Runs with tool:, state:, agent:, and verdict: tokens, enter to focus detail (Failures lists the group's affected runs newest first with their agents), a to jump to the run's ⚒ Runs block on the owning Agents row, v to open the retained log in the pager, y to copy the run id, s to stop the focused live run (same DANGER confirm as the Agents tab), r to run the focused catalog tool at the current project root, and R to reload. The Runs detail reuses the Agents tab Runs block renderer. The pane never settles or reconciles runs; the only writes it offers are the explicit stop and catalog-run flows. The top-bar tools: group opens this pane with all projects, so the pane never shows fewer runs than the chip claims.

_[project]: A sase project is a named unit of work registered with SASE. _[workspace]: A sase workspace is a numbered clone of a project's primary repo, managed by the workspace store and tracked in that project's registry.json. _[agent]: A sase agent is an agent session or a single agent that does not belong to a session. _[prompt]: A prompt is the text SASE hands to a provider to start an agent turn. _[macro]: A macro is a reusable prompt triggered with #name in agent prompts. _[Patch]: A Patch is SASE's local unit of change. _[bead]: A bead is one tracked work item in SASE's git-native issue tracker. _[plan]: A plan is a persisted, approved artifact recording the intent behind agent work. _[TUI]: The TUI is SASE's terminal interface for running agents and reviewing results. _[epic]: An epic is an executable multi-phase plan run with sase bead work. _[gate]: A sase gate is a durable, command-backed request for one user decision, created when work must pause for the user. _[monitor]: A sase monitor is a monitor turn that runs one long command under a detached supervisor, so the command outlives the agent that started it. *[stitch]: A stitch is the lightweight ordered change record inside a Patch's STITCHES: section.