Fleet Management

The Fleet Management view gives you a single pane of glass over every managed runner in your organization. Monitor health, aggregate costs, and scale capacity — all from the dashboard.

Fleet Management requires the Growth plan or above. Free and Starter plans are limited to a single concurrent runner.

Fleet Overview

The fleet table shows every runner with real-time status:

Column	Description
Runner	Name, ID, and current template
Status	`idle` · `running` · `error` · `terminated`
Uptime	Time since last provisioning
Sessions	Active / total session count
Cost (today)	Accumulated LLM spend for the current day
Region	Deployment region (e.g. `eu-central`, `us-east`)

Bulk Actions

Select multiple runners to perform bulk operations:

Start / Stop — Spin up or tear down selected runners
Restart — Tear down and re-provision with the same template
Update Template — Apply a new runner template to all selected runners
Emergency Stop — Immediately terminate all sessions and shut down

Fleet Presets

Fleet presets let you save and recall common fleet configurations:


# CLI: apply a saved preset
curate runners fleet apply my-preset
 
# CLI: list presets
curate runners fleet presets

Presets store:

Number of runners and their templates
Region distribution
Auto-scaling thresholds
Warm pool configuration

Aggregate Health

The fleet health panel aggregates signals across all runners:

Metric	Source
Runner availability	Percentage of runners in `idle` or `running` state
Error rate	Failed sessions / total sessions over the last hour
Avg session duration	Mean execution time across all active sessions
P95 latency	95th percentile LLM response time across the fleet

Health alerts fire when:

Runner availability drops below 80%
Error rate exceeds 5% over a 15-minute window
Any single runner has been in error state for more than 10 minutes

Aggregate Costs

The fleet cost panel rolls up spend across all runners:

Total fleet spend (today / this month)
Cost per runner breakdown with sparkline trends
Cost per session average
Projected monthly spend based on trailing 7-day average

Costs flow from the gateway’s cost tracking system. Each runner’s LLM calls are attributed via the X-CM-Runner-Id header automatically injected by the runner control plane.

Warm Pool

The warm pool keeps pre-provisioned runners ready for instant session start:

Setting	Description	Default
Pool size	Number of warm runners to maintain	2
Template	Runner template for warm instances	`default`
Max idle time	Auto-terminate warm runners after this duration	30 min
Region	Where to place warm runners	Same as fleet


# CLI: check warm pool status
curate runners warm-pool status
 
# CLI: request a warm runner
curate runners warm-pool request --template web-automation

Capacity Planning

The capacity planner helps you right-size your fleet:

Historical analysis — Review peak concurrent sessions over the last 7/30/90 days
Growth projection — Estimate future capacity needs based on session growth rate
Cost simulation — Model the cost impact of adding or removing runners
Recommendation — Suggested fleet size based on your P95 concurrency + 20% headroom

Dashboard Pages

Fleet management spans several dashboard pages:

Page	Path	Purpose
Fleet Overview	`/runners/fleet`	Aggregate status, health, and actions
Fleet Presets	`/runners/fleet-presets`	Save and apply fleet configurations
Command Center	`/runners/command-center`	Real-time operational view
Cost Forecast	`/runners/cost-forecast`	Projected spend and capacity planning
Runner Inspector	`/runners/inspector`	Deep dive into individual runner state