Overview

Multi-cloud Test Playground

Run safe collector-first proofs for AWS, GCP, and Azure. Each test uses bundled provider-style data, passes through the demo collector flow, and shows the stored CloudSight row on the right.

Step 1 Choose a test

Pick a cloud and a test scenario to begin.

Popular tests

View all tests

Step 2 Run the proof

We call the provider, forward to the collector, and store the result.

Step 3 See the result

Live status, captured row, and audit trail.

Quick actions

Live TestRun a single test Captured RowsBrowse stored results Raw JSONView collector payloads Audit TrailView end-to-end logs API ExplorerTest API endpoints

Live Test

Run a single test and verify the result in real time.

1. Select cloud

2. Select service

Cloud Storage

—

4. Configure

Bucket name File name

Test summary

Pick a service

This will

Latest run

Captured Rows

Browse the latest stored CloudSight rows

Browse and search rows captured by the collector and normalized by CloudSight.

Run a test to see the latest matching CloudSight rows.

Time	Cloud	Service	Endpoint	Bucket / Object	Status	Units	Cost

Confirmation

What this run proved

Raw JSON

View the full payload stored by CloudSight

Back to Captured Rows

Select a service on the left and run it.

Audit Trail

End-to-end timeline for the selected row

Back to Captured Rows

Recent calls will appear here. Matching events from the current run are highlighted.

Providers

Manage collectors and cloud provider connections

Current collector status, cloud region, heartbeat, and proof readiness.

API Explorer

Test CloudSight API endpoints

Inspect the selected proof contract and the latest collector response.

Request

Pick a service to see the current proof contract.

Response

200 OK 317 ms

Run a proof to inspect the latest collector result.

CloudSight Client Playbook

Architecture, onboarding, proof, and rollout in one guide

Use this page to explain CloudSight to a new client from zero context. It covers the business problem, platform architecture, collector design, data model, integration steps, security posture, demo flow, go-live plan, and the exact proof points a client should validate before production.

01Collect cloud and AI usage signals close to the source.

02Normalize every signal into one cost telemetry model.

03Prove every dashboard number with audit and readback.

Executive view Benefits Architecture Data model Integrations Security Rollout

01

Executive view

CloudSight is a cost observability platform for cloud and AI spend. It helps leadership know where money is going, helps Finance close the month with better evidence, and helps Engineering find which service, endpoint, region, model, team, or customer is driving cost.

The product is built around one idea: cost data should be observable like reliability data. Every metric should have a source, every row should be traceable, and every report should be generated from normalized telemetry rather than manual spreadsheet stitching.

02

Client outcomes

Visibility: one dashboard for total spend, requests, services, and top cost drivers.
Explainability: endpoint-level usage rows with tags, region, environment, and source reference.
Finance readiness: reports, exports, budget summaries, and allocation views from the same data.
Trust: audit trail from provider signal to collector to CloudSight storage and readback.
Action: identify waste, anomalies, owner teams, and optimization opportunities.

03

Evaluation principle

A serious client should not start by giving broad cloud permissions. They should first run a dry-run proof. CloudSight can replay realistic AWS, GCP, Azure, OpenAI, or internal-service events into the collector contract without creating cloud resources. That proves ingestion, normalization, dashboards, reports, and audit before production access is considered.

Benefits And Features

What clients can use CloudSight for

CloudSight is useful across leadership, Finance, FinOps, Engineering, platform teams, and AI product teams because the same normalized telemetry powers every workflow.

Cloud cost visibility

See total spend, provider split, top services, service share, request volume, active services, and trend changes in one executive dashboard.

Usage explorer

Filter usage by provider, service, endpoint, environment, region, date range, custom range, team, product, and source reference.

AI spend observability

Track OpenAI, Anthropic, Azure OpenAI, model gateways, token volume, model choice, feature usage, and cost per product workflow.

Finance-ready reporting

Generate finance statements, budget reports, allocation summaries, unit economics reports, and exports for monthly review.

Allocation and chargeback

Attribute spend to teams, products, customers, cost centers, environments, subscriptions, projects, or internal business units.

Audit trail and proof

Trace an event from provider signal to collector receipt, CloudSight ingestion, database write, dashboard readback, and report output.

Anomaly and budget alerts

Highlight budget threshold breaches, usage spikes, service drift, unusual endpoint behavior, and unexpected provider growth.

Optimization workflow

Find expensive endpoints, low-value usage, idle infrastructure patterns, provider concentration, and services needing ownership.

Client team Primary use CloudSight features Business result

Leadership Understand spend direction and risk Overview dashboard, spend trend, top service share, budget overview Faster decisions with fewer surprise bills

Finance Close monthly cloud and AI cost review Reports, exports, allocation, audit evidence, custom ranges Cleaner statements and better reconciliation

Engineering Find which service or endpoint caused cost Usage explorer, endpoint table, raw JSON, source references Faster root-cause analysis and optimization

FinOps Govern spend and assign ownership Alerts, budgets, unit economics, team tags, provider filters Accountable cost ownership across teams

Platform Operate integrations safely Collectors, API keys, status page, smoke tests, backup freshness Reliable telemetry pipeline with recovery controls

Reference Architecture

The complete data path

Provider credentials stay in the client's control. CloudSight receives normalized cost telemetry and proof metadata.

Client environment

AWS
Cost Explorer, CloudWatch, CloudTrail, CUR, S3

GCP
Billing Export, BigQuery, Cloud Logging, Storage

Azure
Cost Management, Monitor, Blob, Subscription usage

AI + internal
OpenAI, Anthropic, custom SaaS, product events

Collector layer

Provider adapters
Map raw signals into CloudSight fields

Enrichment
Team, product, region, environment, cost center

Delivery
Signed HTTPS event batches with API key scope

CloudSight platform

API gateway
Authentication, validation, rate control

Postgres model
Usage rows, collectors, reports, audit events

Application
Dashboard, Usage, Reports, Alerts, Allocation

1. Source signal Cloud bill, usage event, model request, endpoint call, bucket object, job, VM hour, or custom business event.

Collect

2. Collector transform Collector converts provider-specific shape into CloudSight's normalized event contract.

Normalize

3. Platform write API validates key, stores usage rows, records audit events, and exposes readback.

Analyze

4. Business output Dashboard, usage explorer, reports, budget alerts, allocation, and unit economics update together.

Important for security reviews: CloudSight does not need broad production admin access. The preferred production pattern is least-privilege read, export, or event-forwarding access held by the collector in the client's environment. CloudSight stores normalized telemetry, not provider master credentials.

Domain Model

Core objects a client should understand

Workspace: The client tenant. It owns users, roles, collectors, reports, API keys, settings, and usage rows.
Collector: A provider-specific adapter that runs in a controlled environment and converts provider signals into CloudSight events.
Usage row: A normalized cost and usage record with provider, service, endpoint, units, cost, region, environment, timestamp, tags, and source reference.
Audit event: Evidence that provider collection, collector dispatch, CloudSight ingestion, verification, or report readback happened.
Report: A finance or operations-ready statement generated from normalized rows, filters, and workspace policy.

Normalized Event Contract

Required and recommended fields

Identityworkspace, provider, collectorName, batchReference, sourceReference

Usageservice, inputEndpoint, outputEndpoint, inputUnits, outputUnits, calculatedCost

ContextregionCode, deploymentEnvironment, project, account, subscription, resourceId

Ownershipteam, product, customer, costCenter, application, ownerEmail

Governancetimestamp, sourceType, schemaVersion, retentionClass, auditCorrelationId

QualityvalidationStatus, currency, pricingSource, confidence, duplicateKey

Integration Blueprint

How a client starts from scratch

The recommended path moves from zero-risk proof to production telemetry in controlled stages.

Stage 1

Discovery and scope

Confirm providers, billing accounts, subscriptions, projects, regions, environments, teams, tags, reports, dashboards, security constraints, and success criteria. This stage defines what "working" means before any collector is connected.

Stage 2

Dry-run proof

Replay saved provider-like events through the CloudSight API and collector contract. Validate captured rows, raw JSON, audit trail, dashboards, reports, allocation, and usage charts without cloud provider writes or charges.

Stage 3

Collector deployment

Run the collector as a container, scheduled job, serverless function, private service, or customer-managed integration. Supply provider credentials locally and issue a scoped CloudSight collector key.

Stage 4

Production validation

Compare CloudSight output against provider console samples, billing export rows, or known service totals. Enable scheduled reports only after source references, units, costs, and tags are accepted by Finance and Engineering.

API Example

Example collector event

{
  "provider": "AWS",
  "collectorName": "aws-prod-cost-collector",
  "mode": "PRODUCTION",
  "batchReference": "aws-batch-2026-08-26T10:00Z",
  "events": [{
    "service": "AWS",
    "inputEndpoint": "api-gateway-request",
    "outputEndpoint": "cloudfront-egress-gb",
    "inputUnits": 24000,
    "outputUnits": 18,
    "timestamp": "2026-08-26T10:00:00Z",
    "sourceType": "CLOUDWATCH",
    "sourceReference": "apigw-prod-orders-10am",
    "regionCode": "ap-south-1",
    "deploymentEnvironment": "Production",
    "tags": {
      "team": "platform",
      "product": "orders",
      "costCenter": "finops-101"
    }
  }]
}

Readback Example

What CloudSight returns

{
  "status": "SUCCESS",
  "stored": 1,
  "provider": "AWS",
  "message": "Usage event accepted",
  "readback": {
    "matched": 1,
    "latestSource": "apigw-prod-orders-10am"
  }
}

DashboardKPI cards and spend trend update.

UsageRow appears by provider, endpoint, and date range.

ReportsFinance exports include the normalized cost.

AuditTrace proves collector receipt and DB write.

Provider Playbooks

What each integration usually collects

Exact permissions depend on the client's environment, but these are the common starting points.

awsAWS

Cost Explorer or CUR for billing truth.
CloudWatch and CloudTrail for service activity and proof.
S3, Lambda, EC2, RDS, API Gateway, CloudFront, DynamoDB signals.
Tags: account, region, product, team, environment, cost center.

GGCP

Cloud Billing Export to BigQuery for invoice-grade data.
Cloud Logging and Audit Logs for traceability.
Cloud Storage, BigQuery, Cloud Run, GKE, Pub/Sub, Gemini signals.
Labels: project, folder, organization, service, team, environment.

AAzure

Cost Management exports and subscription usage details.
Azure Monitor for resource and service metrics.
Blob Storage, VM, Functions, Azure SQL, Cosmos DB, Azure OpenAI.
Tags: subscription, resource group, owner, application, business unit.

AIAI and SaaS

OpenAI, Anthropic, internal LLM gateways, and model routing layers.
Token usage, request count, latency, model, feature, product, user group.
Map tokens and requests to cost per feature, customer, or team.
Useful for chargeback, anomaly detection, and AI product margins.

Security And Governance

Controls a client security team expects

The design should be easy to review because responsibilities are separated clearly.

Least privilegeProvider access starts read-only or event-only. No broad admin role is required for standard telemetry.

Credential separationCloud provider secrets live in the collector runtime, not in the CloudSight UI.

Scoped API keysCollector keys can be issued per provider, rotated, disabled, and tied to a workspace.

Tenant isolationUsers, rows, reports, API keys, collectors, and settings are workspace-owned.

Audit evidenceRows can be traced by sourceReference, batchReference, collector, provider, and timestamp.

Operational controlsHealth checks cover API, database, collectors, ingestion freshness, backup freshness, and smoke tests.

AreaClient ownsCloudSight owns

Provider credentialsCreation, storage, rotation, least privilegeNever requires broad provider admin secrets

Collector runtimeNetwork, deployment, logs, secret storeReference collector, schema, retry guidance

Telemetry contractCorrect source mapping and tagsValidation, storage, reports, readback, audit

Business adoptionOwners, cost centers, report recipientsDashboards, exports, workflow surfaces

Implementation roles

Client cloud admin: creates read-only roles and export permissions.
Client platform engineer: deploys collector and owns runtime logs.
FinOps owner: validates reports, tags, allocation, and budget rules.
CloudSight admin: manages workspace, users, API keys, and onboarding state.

Proof checklist

Collector receives a sample event.
CloudSight stores a usage row.
Raw JSON matches expected provider/service fields.
Audit trail shows provider, collector, ingestion, and verification steps.
Dashboard, usage explorer, report, and allocation views reflect the event.
Dry-run passes before any real provider write action is enabled.

Commercial story

CloudSight is positioned as a cost observability system for engineering and finance together. It reduces surprise bills, shortens finance reconciliation, helps teams understand API-level cost drivers, and creates a repeatable path from raw cloud signals to board-ready reporting.

Client Conversation Guide

How to explain CloudSight in a meeting

Opening

"CloudSight gives you one trusted view of cloud and AI spend by collecting provider signals, normalizing them into usage rows, and turning those rows into dashboards, reports, alerts, allocation, and unit economics."

Risk handling

"We can prove the full data path with dry-run and replay events first. That means no real provider writes and no unexpected cloud charges during evaluation."

Business value

"Finance gets faster statements. Engineering gets endpoint-level cost drivers. Leaders get a dashboard they can trust because every number can be traced back to an event."

Close

"The next step is a one-provider proof: create a workspace, issue one collector key, replay sample events, validate the dashboard, then connect read-only production signals."

Go-Live Plan

Recommended rollout timeline

Day 0Discovery

Confirm providers, billing accounts, target reports, tags, and owners.

Day 1Dry-run POC

Replay sample AWS/GCP/Azure events and verify all CloudSight screens.

Day 2-3Collector setup

Deploy collector in client environment with least-privilege access.

Day 4Validation

Check rows, audit trail, dashboards, reports, allocations, and alerts.

Day 5Production handoff

Enable scheduled reports, health checks, backups, and ownership workflow.

Acceptance Criteria

What "ready for production" means

This is the checklist to close the POC professionally.

Data correctnessSample provider totals reconcile with CloudSight rows within agreed tolerance.

TraceabilityEvery demo row has sourceReference, raw JSON, audit trail, and readback evidence.

UsabilityFinance can generate reports; Engineering can filter by service, endpoint, region, environment, and date.

OwnershipRows carry team, product, project, customer, or cost-center tags where available.

OperationsHealth page, smoke test, backups, restore rehearsal, and collector key rotation process exist.

Cost safetyDry-run and replay flows are proven before any real provider write or paid provider action is enabled.

Common client questions

Do you need write access to our cloud?: No for standard telemetry. The dry-run proof uses sample events. Production should start with read-only or event-forwarding access.
Where do credentials live?: Provider credentials should live in the client collector runtime. CloudSight receives normalized events over HTTPS.
Can this support custom internal services?: Yes. Any internal system can send the same usage event contract with service, endpoint, units, tags, and source reference.
How do we prove numbers are correct?: Use captured rows, raw JSON, readback, audit trail, and reports from the same source reference.

Next actions in this playground

Run a safe live test Review captured rows Inspect raw JSON Trace the audit trail Open API Explorer Open main CloudSight app

Docs

Guides and references for the CloudSight POC

Use these before connecting a production collector or handing the flow to a client team.

Getting started

Overview Quick start Architecture guide

Guides

How tests work Collectors Providers Understanding captured rows

Reference

API reference Data model Status codes

New to CloudSight?

Learn the basics in 5 minutes.

No-charge simulator

Prove CloudSight without touching cloud providers

Pick a provider and replay a realistic saved event through the CloudSight collector path. This validates ingestion, normalization, database write, dashboard readback, raw JSON, and audit trail without AWS, GCP, Azure, OpenAI, or OCI API calls.

0provider APIs called

4supported providers

1collector contract

Fulldashboard readback

What runs

Build a saved provider-like event.
Send it to the CloudSight collector route.
Normalize provider, service, endpoint, units, tags, and cost.
Write one row into the CloudSight workspace.
Verify dashboard, usage, report, raw JSON, and audit readback.

What never runs

No S3 object creation
No GCS upload
No Azure Blob write
No OpenAI model call
No OCI provisioning

Choose a provider, then run a no-charge proof.

Review captured rows Inspect raw JSON Trace audit trail Open API Explorer Open product status

Environment

Manage your playground environment

Rows in this environment are demo-only and safe to reset.

Environment name Region Data retention Default cloud context

Help / Support

We are here to help you succeed

When a proof fails, use the audit trail first, then review the collector response and CloudSight result details.

Documentation

Browse guides and API reference.

Community

Join the implementation discussion.

Contact support

Get help from the team.

Status

All systems operational.