Resources

Frequently asked questions

Straight answers to what buyers ask before a demo: what Jetscale does, the access it needs, how pricing works and how fixes reach production.

Last reviewed

Product and how it works

Jetscale finds inefficiencies and savings across AWS, Azure and Google Cloud and hands your team each fix as code. It connects with read-only access, scans compute, storage, databases and network in every connected account, and ranks each opportunity by its dollar value, with the evidence behind it. Each finding becomes PR-ready Terraform, or a step-by-step ClickOps runbook, that your engineers review and merge. Jetscale is built by Jetscale AI, a Montreal AI lab founded in 2025 that creates agentic FinOps products. See how the platform works.

A FinOps dashboard shows where the spend goes and leaves the fix to your team. Jetscale goes further. It turns each opportunity into PR-ready Terraform your engineers review and merge through their normal workflow, with the evidence behind it, then tracks each saving until it is realized on your bill.

The native tools are a good free baseline, and Jetscale reads their findings too. Its read-only AWS role includes Compute Optimizer and Cost Optimization Hub. But native recommendations look at each resource on its own, in one cloud, over a short window, mostly propose reservations and basic cleanup, and leave the change to your team. Jetscale goes deeper. It works across resources and across all three clouds at once, using up to a year of telemetry with peaks kept, plus tags and environment, so it catches what a one-resource view misses. Then it hands your team each fix as PR-ready Terraform.

A general assistant only knows what you paste into it. It doesn't see your bill, a year of usage history or how your accounts fit together, so its fixes stay generic. Jetscale's models are trained specifically on cloud infrastructure and optimization, and they analyze your own estate: up to a year of telemetry across every connected account, with each finding ranked by its dollar value. Each fix comes as PR-ready Terraform your team reviews. And you don't have to choose: ChatGPT, Claude or Cursor can query Jetscale's findings through its MCP server.

Jetscale adds to your FinOps team rather than replacing it. The team keeps the practice: priorities, budgets and the conversation with finance. Jetscale makes that job easier. It works between their reviews, keeps scanning every connected account and turns each finding into PR-ready Terraform an engineer can merge, so nobody has to translate a recommendation into a ticket. It also works with the tools the team already uses. See which tools Jetscale connects to.

No. Monitoring shows signals and incidents. Jetscale uses those signals (plus cost and configuration context) to generate optimization recommendations and fixes.

Jetscale looks across compute, storage, databases, containers and network. On AWS that covers EC2 and EBS rightsizing, S3 lifecycle tiering, RDS, Redshift and DynamoDB sizing, ECS and EKS resource requests, Spot moves and missing VPC endpoints. On Azure it covers virtual machines, managed disks, Blob tiering, AKS and reservation coverage, and on Google Cloud, Compute Engine, persistent disks, GKE, BigQuery, Cloud Storage tiering and committed use discounts. Each finding comes with a recommended action, such as rightsize, move to Graviton or Spot, scale in or delete, and a dollar estimate.

Yes. Jetscale covers AI token spend on Amazon Bedrock, analyzing model invocation costs alongside the rest of your AWS estate, and data warehouse usage on Google BigQuery. That puts AI spend in the same costed view as compute and storage. For Azure OpenAI or Google Vertex AI, talk to us about your setup.

Autonomous optimizers change your environment for you, so they need write access to your cloud, often through an agent running inside it, to apply whatever they decide. In an enterprise that's a hard trade: one bad change in production can cost more than it saves, and a vendor holding the keys to your infrastructure is a risk in itself. Jetscale is built for that setting. It stays read-only and hands your team each change as code, with the evidence behind it, so you decide what changes and when. No vendor gets write access to your cloud, and nothing changes until your engineers merge it.

Managed Kubernetes on all three clouds: Amazon EKS, Azure AKS and Google GKE. Jetscale tunes pod resource requests to real usage and recommends node pool and cluster rightsizing, and on AWS it does the same for ECS tasks. EKS Auto Mode and Fargate are coming before the end of 2026, and on-premises clusters and OpenShift are on the roadmap. Each change ships as code your team reviews, like every other fix.

Up to a year of telemetry, with peaks kept rather than averaged away, along with tags, attached resources and the effort each change takes. Sizing against peaks instead of averages is what protects a rightsized resource at month-end or during a seasonal spike.

Savings and pricing model

Self-funded, with guaranteed ROI. Jetscale charges a fixed annual license fee, sized on the cloud estate it covers, and verifies savings against your actual bill. If verified savings come in below the fee, we credit the difference. Net of the savings, the most you'll ever pay is $0.

No. Jetscale charges a fixed annual license fee, sized on the cloud estate it covers. It doesn't move with your monthly bill and isn't a share of your savings, so finance knows the cost for the year up front.

If the savings Jetscale verifies on your cloud bill come in below the fee, Jetscale credits the difference. The terms are set out in your agreement.

Typically 15 to 35% of the cloud spend Jetscale covers, depending on how your estate is built. The free trial, or the free assessment on AWS, gives you your own number. Once you're live, Jetscale reports savings identified and savings realized separately, and realized savings are verified on your bill. Our customer stories show what three companies found, and what one has realized so far.

Jetscale tracks every saving from the moment it is found to the moment it shows up on your bill. Each one sits in one of four states, Realized, Open, Deferred or Discarded, with a realization rate and a week-by-week trend, so savings identified and savings realized are never confused. Realized savings are verified against your actual cloud bill, and those verified savings are what the pricing guarantee counts.

A cleanup resets the meter. It doesn't stop it. Every year brings new instance generations that make last year's right size wrong, commitments that expire and need replanning, new projects that launch oversized, and AI spend that wasn't there before. Jetscale keeps scanning every connected account, so each of these turns into a new fix as your estate changes. Capacity planning keeps your commitments sized to what you actually run, and AI spend sits in the same costed view as the rest of your bill.

Security and access

No. Jetscale only reads from your cloud and never holds write access to it. Every connection uses a dedicated, least-privilege identity limited to read operations, with no standing write access and no shared or long-lived credential. Each fix arrives as PR-ready Terraform, or as a step-by-step ClickOps runbook, and nothing changes until one of your engineers merges or applies it. Connecting GitHub or Azure DevOps so Jetscale opens the pull requests itself is optional. The permissions for each cloud are listed below.

Read-only permissions, granted to an identity you create and control. On AWS, a cross-account IAM role that Jetscale assumes with sts:AssumeRole and a unique External ID. Its policy covers Cost Explorer, Cost Optimization Hub, Compute Optimizer, EC2, RDS, S3, CloudWatch, EKS and related services, using only Describe, Get, List and Export actions. On Azure, an App Registration with the Reader, Cost Management Reader and Monitoring Reader roles at subscription scope, or an optional custom role that also excludes listing or regenerating storage account keys. On Google Cloud, a dedicated read-only identity scoped to cost, billing and resource-inventory data. Security and access has the full detail.

Telemetry and metadata only: CPU and memory usage, tags, resource configuration and billing data. Jetscale never reads the content of your workloads, such as the files in your storage or the rows in your databases, and it doesn't collect IP addresses or hostnames. You can also exclude sensitive tags.

Yes. Jetscale AI is SOC 2 Type 1 and Type 2 compliant. Your security team can request the audit reports under NDA through the Jetscale AI Trust Center.

Yes. Jetscale's team walks through the exact IAM policies, RBAC role definitions and connection architecture with whoever on your side signs off. Every new connection also goes through a validation step before it goes live, which confirms the permissions are scoped correctly and nothing more. Secrets that have to change hands, such as an External ID or a client secret, are generated by the side that controls them and encrypted before they are sent, and the password to open them travels by a separate channel.

Yes. Each person gets a role per business unit: No Access, Viewer, Operator or Admin. The same person can be an Admin in one unit and have no access at all to another, which fits companies that run several teams, subsidiaries or environments under one Jetscale account. Admins add or remove people at any time, removal takes effect immediately, and Jetscale never asks an admin to see or set someone else's password.

None is required. Jetscale produces each fix as PR-ready Terraform that your team can open as a pull request itself. If you'd rather Jetscale open the pull requests, you can connect GitHub or Azure DevOps through the provider's own connected-app (OAuth) authorization, never through a pasted static token. Jetscale then uses that connection to open pull requests with its recommended changes against your infrastructure-as-code repositories. You can see what is connected and disconnect it at any time from the screen where you set it up, and your team still reviews and merges every pull request.

Clouds, tools and code

AWS, Microsoft Azure and Google Cloud, all three at the same depth and in one costed view. Accounts, subscriptions and projects from every provider sit side by side and can be grouped into business units, so a multi-cloud estate is analyzed and prioritized as one. See Jetscale for AWS, Azure and Google Cloud.

No. Jetscale delivers each fix either as PR-ready Terraform or as a guided ClickOps runbook with the exact console URLs, so a team that changes resources by hand can apply it too. Either way, the fix comes with a structured plan: what changes, the expected impact and the safest rollout path. Other infrastructure-as-code formats, such as Bicep, CloudFormation, Pulumi and CDK, are on the roadmap for early 2027.

Terraform, today. Jetscale writes each fix as PR-ready Terraform, which your team opens as a pull request, or Jetscale does if you connect GitHub or Azure DevOps. Bicep, CloudFormation, Pulumi and CDK are on the roadmap for early 2027. Until then, for a resource you don't manage in Terraform, the same fix comes as a guided ClickOps runbook.

Jetscale reads from your AWS, Azure and Google Cloud accounts. Through optional integrations, it opens pull requests in GitHub and Azure DevOps, so fixes land in the CI/CD and change-management process you already run, and it connects to Slack. Its MCP server lets ChatGPT, Claude, Cursor or any MCP-compatible client query its findings. More integrations are coming in early 2027. See all integrations.

Yes. Jetscale ships a Model Context Protocol (MCP) server, so ChatGPT, Claude, Cursor or any MCP-compatible client can query findings, cost data and recommendations directly. MCP access inherits the connecting person's Jetscale role and business-unit scope, so an agent never sees more than that person could. More on MCP.

Yes, through a custom connector. Jetscale has no native Power BI connector, but it has built custom connectors for clients that bring its findings and cost data into Power BI or a data lake. Your own AI agents and tools can also query Jetscale's findings through its MCP server.

Spot and commitments

Jetscale only recommends moving a workload to Spot after OptimizeSpot has simulated its eviction risk against that workload's own usage history. If the move isn't safe, it isn't recommended. The same check runs for Spot on AWS, Spot VMs on Azure, and Spot and preemptible VMs on Google Cloud, and the move ships as code your team reviews like any other change.

A commitment broker lowers the rate you pay by buying reserved instances and savings plans for you. Jetscale lowers what you need to buy in the first place: right-sizing, storage, idle resources and network, across AWS, Azure and GCP. The two stack, and your commitments then cover a smaller baseline. By the end of 2026, Jetscale will also bring full commitment and capacity planning into the platform, so you can size the baseline and plan the commitments that cover it in one place.

Yes. Jetscale's capacity planning recommends which reservations and savings plans to buy, sized on what your workloads use after rightsizing, so you commit to what you will actually run. It also checks the commitments you already hold against actual usage, so steady-state workloads aren't left on on-demand rates and you aren't over-committed. Jetscale stays read-only, so it recommends each purchase and your team makes it.

Getting started

Yes, three ways. You can start a free 7-day trial today on one cloud account, for up to $1,000 in savings. On AWS, you can also get a free cloud assessment through AWS Marketplace: you subscribe at no cost, connect a read-only IAM role, and get a ranked list of savings opportunities with monthly estimates. For a free assessment on Azure or Google Cloud, or a proof of concept on your own estate, contact us.

About 15 minutes to connect an account, and the first findings within hours. Connecting means creating a read-only identity in your cloud, such as a cross-account IAM role on AWS, which Jetscale validates before it goes live.

Your engineers. Jetscale hands over each fix as PR-ready Terraform or a ClickOps runbook, and opens the pull request itself if you've connected GitHub or Azure DevOps. Your team reviews, merges and deploys the change through the CI/CD and change-management process you already run. Jetscale never applies a change itself, because it never has write access to your cloud.

Each recommendation carries what an engineer needs before touching production: the implementation effort, whether a restart is required, whether the change is rollback-safe, the exact resource IDs affected and the evidence behind it. The fix ships as a structured plan covering what changes, the expected impact and the safest rollout path. Your team reviews it like any other change, merges it only when satisfied, and confirms the impact after rollout.

From inside the platform. Linking another AWS account, Azure subscription or Google Cloud project is a guided "Link a new cloud account" flow, with no separate setup project. Each account can be grouped under a business unit with its own name, description and owner, and people get their roles per business unit.

About Jetscale AI

Jetscale AI, a Montreal AI lab founded in 2025 by Mehdi Merai, CEO, and Gabriel De Lisi, CTO, who have built enterprise AI systems since 2014. In February 2026 it closed an oversubscribed $5.4 million seed round co-led by the BDC Seed Venture Fund and the Diagram ClimateTech Fund, bringing its total funding to $6 million. About Jetscale AI.

Yes to both. Jetscale AI - Free Cloud Assessment is listed on AWS Marketplace and free to subscribe, and Jetscale AI is a General Member of the FinOps Foundation. Partners and affiliations links to both.

08Get started

See your number before you spend a dollar.

Read-only connection. Evidence-backed findings. PR-ready fixes. Self-funded, with guaranteed ROI.