Cloud signals +43.8%

Turn cloud data into decisions that hold up.

Flux connects cost, usage, inventory, and governance into one continuously updated intelligence layer for cloud operations.

Read-only, managed identity Connected in minutes
flux · overview
The Flux overview dashboard showing estate cost, utilization, and opportunity charts
Built for teams that need one cloud truth.
The reconciliation tax

Four teams, four spreadsheets, four different totals.

Cost lives in one export, inventory in another, utilization in a monitoring tool, and governance in a portal blade. By the time anyone reconciles them, the estate has changed. Flux collapses that into a single governed layer that updates itself — and tells you when it couldn't.

Cost that reconciles

Actual and amortized spend, checkpointed daily history, and charge-level detail modelled on FOCUS, the open FinOps billing standard — cross-checked across sources so the total on the dashboard is the total on the invoice.

Inventory that remembers

Every inventory snapshot is retained, so "what changed?" is an exact diff between two points in time — not an educated guess from a current-state view.

Utilization that's honest

Platform metrics and LogicMonitor reconciled per resource, with explicit coverage states. Missing telemetry is never quietly classified as idle.

Onegoverned source behind every view
Read-onlyno cloud mutation, no stored secrets
Everynumber traceable to a validated snapshot version
Measuredsavings from real cost deltas, not estimates
What you get

Nine things that go to work on day one.

Every one is live in the product today, and every one links to its documentation — including the parts that explain where it stops.

Allocation

Showback that survives a challenge

Spend attributed to cost centres by tag, shared infrastructure prorated in proportion to real usage, and the unallocated remainder kept visible as its own number rather than quietly absorbed.

Allocation coverage as a headline metric →
Anomalies

Know what moved, and what moved it

Matching-weekday baselines that understand your weekly rhythm, with service and resource contributors ranked by absolute change — so "spend rose 22%" becomes "this cluster did it."

Detection, contributors, review workflow →
Right-sizing

Savings you can defend in review

Findings computed from reconciled telemetry with explicit coverage states, split into what is actionable now versus what still needs evidence — and never calling an unmonitored machine idle.

Evidence, confidence, valuation →
Outcomes

Savings measured, not estimated

Track a recommendation from accepted to implemented, then measure what it actually saved from cost data after the change. The number you report upward is the one that survives audit.

How realization is measured →
Commitments

Find the capacity you paid for

Per-commitment utilization showing used against unused, coverage across eligible usage, and realized discount against list price by service — computed from charge-level billing data.

Coverage, utilization, discounts →
Change

An exact answer to "what changed?"

Creates, deletes, resizes, re-tiers, re-tags and moves as a precise diff between two snapshots, with before and after values — plus detection on change volume itself.

Inventory and change tracking →
Budgets

Targets with an honest projection

Monthly budgets per subscription or estate-wide, with on-track, at-risk and over status — and a projection method stated plainly enough that a finance partner can reproduce it.

Budgets and unit economics →
Forecast

A forecast that admits uncertainty

Calendar-month projection with robust confidence bounds and a published accuracy measure, built on matching-weekday history and honest about unfinalized billing days.

Cost reporting and forecasting →
Flux Intelligence

Ask in plain language, get the receipts

An AI agent that plans its own investigation across governed data, separates fact from inference, links every claim back to the page that proves it — and tells you what it could not cover instead of filling the gap.

Asking questions, and how it is bounded →
Architecture

Collection never blocks a reader.

Flux separates the system that collects from the one that writes and the ones that read. A single writer owns all analytical mutation and publishes validated, checksummed, versioned snapshots. Each serving instance holds its own read-only copy — so ingestion can run flat out while your dashboard stays instant.

Nothing reaches a screen unvalidated, and a bad dataset rolls back by moving a pointer.

Explore the architecture →
SOURCES WRITER SNAPSHOT SERVING Resource Graph Cost + FOCUS Advisor Telemetry Single writer idempotent staged applies DuckDB analytics lineage kept Validated checksummed versioned immutable instance A own copy instance B own copy PostgreSQL control plane leases · claims · retries · publication ledger
Flux Intelligence · agentic

An AI agent that investigates, and cannot act.

Flux Intelligence plans its own investigation — chaining governed, authenticated reads until it can answer — over a boundary it cannot cross. Never a database connection, never generated SQL, and no tool that writes anything anywhere.

The engines behind the numbers stay deterministic; the agent retrieves and explains them. When Flux has no engine for something, it says so rather than inventing a figure that looks authoritative.

Meet Flux Intelligence →
ASK FLUX · READ-ONLY
YOU
Why did spend jump last Tuesday?
FLUX INTELLIGENCE
investigate_cost_changeget_cost_anomalies
The increase concentrates in a single subscription, driven by a compute meter that rose well above its matching-weekday baseline. Contributors are ranked by absolute change with charge-level evidence.
Partial coverage: the two most recent days are still unfinalized and excluded.

The constraints are the point

No autonomous cloud mutation. No arbitrary SQL. No secrets in the codebase. Readers cannot start or configure synchronization. Successful source data survives partial failures, and a partial total is never presented as an estate-wide one. These are enforced in code and covered by tests — not stated as intentions.

See it against your own estate.

Connect a subscription read-only and watch the first snapshot publish. Most teams start with allocation coverage or a cost spike nobody can explain.