MONEY SAVED
Total saved
$0.00
0% of est. cloud cost
Cloud cost (no cache)
$0.00
est. at Sonnet rates
Actual cost
$0.00
tokens spent via pipeline
Cache hit rate
0%
0% resolved locally
DEVELOPER
PRODUCTIVITY
Dev time saved
0s
wait time eliminated
Dev cost saved ($80K)
$0.00
junior/mid developer
Dev cost saved ($250K)
$0.00
senior/staff engineer
Avg response time
0ms
vs ~5000ms cloud
COST FORECAST
Monthly projection
—
based on current usage
Avg cost/query
—
across all tiers
Queries/day
—
rolling average
Days to $10
—
at current spend rate
ROI TIMELINE
Savings
—
saved
vs. cloud cost
Local
—
local
cache + local = $0
Efficiency
—
/100
routing score
Cumulative savings vs. cost
TIER
DISTRIBUTION
Queries by tier
2.5D Cache Shard View
QUERY
HEATMAP
Queries by day — click a cell for details
Less
More
CACHE +
OUTPUT
Cache memory (0 docs)
code
research
deploy
creative
note
other
Output value
Queries processed0
Lines of output ▸0
Total tokens produced0
Cost per line$0.000
Token cost avoided—
QUALITY &
COMPLEXITY
Quality metrics
Cache escalation rate
—
Avg cache similarity—
Local resolution rate—
Routing efficiency—
Escalation = dev rejected cached answer and re-queried cloud
Token economics
Cost / token (actual)—
Avoided / token—
Worst case / session—
Worst case / year—
Worst case = 1M tok Opus 4 Extended runaway session
ENTERPRISE
PROJECTION
500 developers · 250 work days/yr · 6h active agent time/day
Per-dev API spend (no cache)—
Annual cloud spend (no pipeline)—
Observed cache+local hit rate—
With muLLM (projected)—
Projected annual savings
—
Assumptions & methodology
Dev daily spend (low)—
Dev daily spend (high)—
Opus-heavy sessionsup to $50/session
Cache hit rate (warm)40–60%
Local tier cost$0.00/query
Cost modelSonnet 4
Per-dev daily spend computed from 150–400 queries × selected
benchmark model rate. Source: typical active Claude Code / Cursor
usage at 6h/day. Cache savings compound with team size — shared
query patterns reduce per-dev cost. All projections are estimates
— actual savings depend on your query mix and cache hit rate.
QUEUED MODULES
Anomaly Detection
Real-time alerts when cost patterns spike
Cost Stats API
Live! /api/stats — per-tier, per-day, per-model breakdowns
Savings Velocity
Live $/hour accumulation rate
Shareable Summaries
Live! Click SHARE button in topbar
Screenshot Export
Live! Click CAPTURE button in topbar
LIVE QUERY
STREAM
(dev mode)
CACHE (free)
LOCAL (free)
MULTI (free)
CLOUD CHEAP ($)
CLOUD FULL ($$)
ROUTING REPLAY
Loading routing decisions…
— · Fetched from
/api/routing/replay