| name | memfleet-fleet-status |
| description | Always use as the once-per-session opener and as the periodic dashboard refresh. Returns active intents, open subscriptions, episode counts, and conflict counts split by class. Triggered by: 'how busy is the fleet', 'are there open conflicts', 'is anyone working on this repo', pre-release coordination check, leader pre-flight before a coordinated wave, periodic ops dashboard. Do not skip this call before launching a multi-symbol refactor — high class-B/C counts mean the wave will collide. Do not use it as a per-symbol query; for that use get_node_state. |
Overview
One-shot dashboard for the broker's coordination state across all repos (or one repo if scoped). Cheap enough to call every few minutes; designed to be the session-opener after memfleet start.
Returns four things you need for any coordinated decision:
- Active intents (who is working on what, right now)
- Open subscriptions (who is watching what)
- Episodes recorded in the last window (recent activity volume)
- Conflict counts split by class (A/B/C — the health signal)
Quick Reference
| Tool | Purpose |
|---|
fleet_status | Cross-repo or repo-scoped coordination snapshot |
Parameter types: MCP parameters are strictly typed. window_ms MUST be a JSON number, not a quoted string. repo_id is a string; omit to get a cross-repo rollup.
Steps
1. Pick the window
window_ms controls how far back to count episodes and conflicts. Sensible defaults:
| Use case | window_ms |
|---|
| Session opener | 1_800_000 (30 min) |
| Pre-coordinated-wave check | 300_000 (5 min) — only the freshest signal matters |
| Periodic dashboard refresh | 60_000 (1 min) |
| Daily ops review | 86_400_000 (24 hr) |
2. Call fleet_status
mcp__memfleet__fleet_status({
repo_id: "my-repo", // omit for cross-repo
window_ms: 1_800_000
})
Parameters:
repo_id — string, optional. Scope to one repo. Omit for fleet-wide rollup.window_ms — integer, optional. Lookback window for episode and conflict counts. Default 1,800,000 (30 min).
Success criteria: the broker returns active intents, subscriptions, episode count, and conflict counts.
3. Interpret the response
| Signal | Meaning | Action |
|---|
active_intents == 0 | Fleet is idle in this window | Safe for large refactors |
active_intents is high but conflict_count_C == 0 | Healthy busy fleet | Proceed; expect normal Class B arbitration |
conflict_count_C > 0 | Active blocking conflicts | Read each via query_episodes with conflict_class: "C" before adding work |
conflict_density on a module above the team's threshold | That module is a battleground | Avoid it OR coordinate via memfleet-fleet-coordination |
episodes_in_window very high vs baseline | Burst of activity | Likely a coordinated wave in progress; check for shared correlation tags |
subscriptions_open == 0 and you expected watchers | Phase-1 stub or watchers crashed | Don't rely on push; poll query_episodes instead |
Decision Points
| Situation | Action |
|---|
| Session opener | Call once with window_ms: 1_800_000; cache the result for the rest of the session |
| About to launch a coordinated wave (leader role) | Call with window_ms: 300_000; if any class-C is open, abort the wave |
| Periodic dashboard | Call with window_ms: 60_000 every few minutes; show deltas |
| Need per-symbol activity | This tool is the wrong shape — use get_node_state instead |
| Need to filter by intent_kind | Combine with query_episodes(intent_kind=...); fleet_status is unfiltered by design |
| Cross-repo rollup | Omit repo_id |
Common Mistakes
| Mistake | Reality |
|---|
Using fleet_status for per-symbol checks | fleet_status is a global rollup. For per-symbol, use get_node_state. |
| Skipping it before a coordinated wave | A wave on top of an open Class C produces unresolvable state. Always pre-check. |
Quoting window_ms as a string | Must be a JSON number. |
Treating subscriptions_open == 0 as a bug | Phase-1 subscriptions are stub. Poll instead. |
| Calling it in a tight loop | Designed for periodic refresh, not continuous polling. Every-few-minutes is fine; every-second is wasteful. |
Phase-1 Note
fleet-cli is the interactive face of the same data. Use fleet_status inside agents and fleet-cli for human eyes. Both go through the same broker tool; they cannot disagree.
Skill Priority
This is a session-opener + periodic ops tool. Pair with:
