How the Databricks cost tools differ
This page puts Databricks' own tools, 19 other products and ours side by side. For each one you can see what it does, how it connects to your account and how it is priced.
We make ClusterBite, so read this with that in mind. Everything here comes from each vendor's own public pages as we read them on , and every entry links to its sources. Where a price isn't published we say so. Products and prices change, so check the links before you decide. If you find a mistake, tell us and we'll fix it.
At a glance
"Shows costs" means dashboards and alerts. "Suggests fixes" means it tells you what to change. "Changes things itself" means it can change settings in your account for you. "Partly" means only some of it. "Price published" is Yes when every option has a price on the vendor's own pages, Partly when only some do (a starting price, a free tier or some plans), and No when none do.
| Tool | Shows costs | Suggests fixes | Changes things itself | Pricing model | Price published |
|---|---|---|---|---|---|
| Databricks built-in tools | Yes | Partly | No | Included with Databricks | Yes |
| Lakesight | Yes | No | No | Monthly, by workspaces | Yes |
| Vantage | Yes | No | No | Tiers by tracked cloud spend | Partly |
| CloudZero | Yes | No | No | Subscription, by cloud spend | No |
| LakeSentry | Yes | Yes | No | Flat monthly plans | Yes |
| Flexera Data Cloud Optimization | Yes | Yes | No | Not public | No |
| PointFive | Yes | Yes | Yes | Not public | No |
| Unravel Data | Yes | Yes | Yes | By DBUs monitored | Partly |
| Lakemine | Yes | Yes | Yes | Share of savings, or license by DBUs | No |
| ClusterBite | Yes | Yes | No | Share of savings, three months | Partly |
| Keebo | No | No | Yes | By DBUs saved, or subscription | Partly |
| Espresso AI | No | No | Yes | Share of savings | Partly |
| Zipher | Yes | No | Yes | Pay as you go, or enterprise plan | No |
| Altimate AI | Yes | Yes | Yes | Per account | No |
| Sedai | No | Yes | Yes | Per vCPU managed, yearly | No |
| Revefi | Yes | Yes | Yes | Percentage of spend | Partly |
| Capital One Slingshot | Yes | Yes | Yes | Not public | No |
| SELECT | Yes | Yes | Yes | Custom quote | Partly |
| Seemore Data | Yes | Yes | Yes | Not public | No |
| DoiT PerfectScale for Databricks | Yes | Yes | Yes | Not public | No |
| Usage.ai | No | Yes | Yes | Share of savings | No |
Tool by tool
Databricks’ own tools
Databricks built-in tools
Included with Databricks
System tables hold your billing and usage history. Usage dashboards chart spend, budgets email you when spend passes a threshold, compute policies limit what clusters people can create, and tags split costs by team. The Cost page in Governance Hub (in beta) adds cost drivers and some recommendations. Suits everyone, as the starting point: seeing where the money goes, splitting it by team and getting alerts.
Access: Built into your Databricks account. Account admins create budgets and import the dashboards; people need read access to the billing system tables to use them.
Price: Part of Databricks, with no separate product to buy. The queries and dashboards run on your own compute.
Sources: cost management tools, system tables, usage dashboards, budgets, compute policies, Governance Hub
Cost reporting
Dashboards, alerts and reports. None of these suggest fixes.
Lakesight
Monthly, by workspaces
Databricks cost dashboards: machine (VM) and DBU cost for every job run, live tracking of running jobs and clusters, email alerts when jobs fail or run too long (including runs started from Azure Data Factory or Airflow), and scheduled cost reports.
Access: A workspace URL plus an access token or OAuth service principal. Read operations only, with no access to your cloud subscription.
Price: $49 a month for 3 workspaces and up to 20 team members; extra workspaces $29 a month each. 14-day free trial.
Vantage
Tiers by tracked cloud spend
A cloud cost platform with a Databricks integration: DBU charges joined with the AWS or Azure machines underneath, spend by workspace, cluster, job or user, alerts on spikes and budget overruns, and forecasts. It can apply your own discount rates. Suits smaller teams that want Databricks and cloud costs in one report at a low published price.
Access: Read-only: a Data Reader permission and a dedicated serverless SQL warehouse to query the system tables. No write access.
Price: Free for up to $2,500 of tracked cloud spend. Pro $30 a month (up to $7,500) and Business $200 a month (up to $20,000), each with a 14-day trial. Enterprise is custom. Limits count all the cloud spend you track, not just Databricks.
Sources: Databricks, pricing
CloudZero
Subscription, by cloud spend
A cloud cost platform that brings Databricks spend in next to the rest of your cloud bill: cost by team, product and customer, unit costs, and alerts on unusual spend sent to the owners.
Access: A service principal that reads the billing and compute system tables through a SQL warehouse; CloudZero says all its connections are read-only. Its guide covers Databricks bought directly on AWS; Databricks bought through Azure shows up in the Azure billing connection.
Price: Not public. One subscription with all features, quoted on the scale of your cloud and AI spend.
Sources: Databricks, pricing, docs: connecting Databricks
Findings for your team to act on
These list what to change. Unravel, PointFive and Lakemine can also make some changes once you allow it.
LakeSentry
Flat monthly plans
Databricks cost monitoring: cost breakdowns and chargeback, alerts on unusual spend, and optimization insights ranked by estimated saving.
Access: A read-only service principal that can read the system tables. Hosted, as a Databricks App inside your workspace, or self-hosted in your own network (Docker or Kubernetes).
Price: Free plan (1 user, 3 months of history). Standard $499 a month and Pro $849 a month, billed yearly. All plans include unlimited workspaces.
Sources: home, pricing, hosted setup, private cloud
Flexera Data Cloud OptimizationFlexera (formerly Chaos Genius)
Not public
Databricks and Snowflake cost by job, cluster, workspace and warehouse inside Flexera One, with chargeback, budgets, unusual-spend alerts and recommendations on idle and oversized compute. Flexera bought Chaos Genius in January 2026; this is that product. Suits companies already on Flexera One, or that want Databricks, Snowflake and cloud costs in one FinOps tool.
Access: A service principal that can read the system tables, using a SQL warehouse in each region. The free assessment needs nothing installed and uses read-only metadata access.
Price: Not public. A free 14-day Databricks assessment is offered.
Sources: Data Cloud Optimization, docs: Databricks connection, Flexera acquires ProsperOps and Chaos Genius (press release)
PointFive
Not public
A cost platform for cloud, data and AI. For Databricks: cost visibility and findings across compute, storage and model serving, sent to Slack, Jira, Teams or ServiceNow, with optional automated fixes.
Access: A notebook you run once creates a read-only service principal for finding waste. Automated fixes use extra permissions you grant, approved per change, once per playbook, or not at all.
Price: Not public. Its pricing link leads to a contact form.
Sources: Databricks, FAQ, contact (pricing)
Unravel Data
By DBUs monitored
A cost and performance platform for Databricks (and Snowflake and BigQuery): cost breakdowns, chargeback and budgets, advice on jobs and clusters, and automated actions. Suits larger data teams that want cost, performance and job tuning in one platform.
Access: Connects to your Databricks workspace and reads the billing and pricing system tables, along with job and cluster details.
Price: Free to start. Paid use is priced on the DBUs it monitors each month, paid yearly or as you go. Rates are not public.
Sources: Databricks, pricing, docs: system tables setup
Lakemine
Share of savings, or license by DBUs
Runs inside your Databricks account and checks it daily with 17+ detection engines, ranks fixes by dollar impact, and checks the savings against your actual invoice. Spend by user, team and workload, alerts, and export to Apptio Cloudability. It can apply approved quick wins itself, with rollback.
Access: Read-only access to billing, performance and configuration metadata, deployed inside your Databricks account; Lakemine says nothing leaves it. Supports AWS, Azure and GCP.
Price: Not public. Two models: pay as you save (a percentage of audited savings, invoiced monthly) or a license scaled to your active DBU use.
Sources: home and pricing, introducing Lakemine
ClusterBiteClusterBite (us)
Share of savings, three months
A cost dashboard plus a ranked list of fixes, each with an estimated monthly saving. Your team decides and makes the changes; we measure what each one saved. Azure Databricks today; AWS and GCP are next. Suits Azure Databricks teams that want to approve every change themselves and pay a share of measured savings.
Access: Read-only: a service principal (or an access token) that can read the Databricks system tables through one SQL warehouse. We host it, or we install the app (a Docker container) inside your network.
Price: 30% of what each change saves, for the three months after it is made, then nothing. The hands-on review is quoted.
Sources: pricing, service agreement
Tools that make the changes themselves
Keebo
By DBUs saved, or subscription
Adjusts Databricks SQL Serverless warehouses (size and idle time) automatically, within limits you set, and reports what it saved. Its docs say the Databricks support is in preview.
Access: An admin creates a service principal for Keebo. It works from performance metadata, read through the Databricks REST API and system tables.
Price: Pay as you go from $0 a month, charged on the DBUs it saves (1 DBU saved = 1 Keebo credit; the credit price is not on the page), or a custom enterprise subscription. 14-day free trial.
Sources: Databricks, pricing, docs: Databricks setup
Espresso AI
Share of savings
Machine-learning agents that scale, schedule and tune Databricks SQL warehouses and other workloads automatically. It also covers Snowflake. Suits teams that want SQL warehouses tuned automatically and prefer to pay out of the savings.
Access: An admin runs a setup script that creates a service principal allowed to manage SQL warehouses and read the system tables. Espresso says it only looks at metadata, never your data.
Price: A share of measured savings: 40% of each month’s savings, or 36% of estimated yearly savings paid upfront (refundable), or custom enterprise billing.
Sources: Databricks launch, pricing, docs: Databricks onboarding
Zipher
Pay as you go, or enterprise plan
Picks cluster settings for Databricks jobs automatically, learning from past runs, with its own autoscaling and spot instance handling. It also shows Databricks and cloud costs together.
Access: Installed into your environment (Zipher says in about five minutes) and changes cluster settings for you. The exact permissions are not on its public pages.
Price: Not public. Pay as you go with a 30-day free trial, or an Enterprise plan with volume and long-term commitment discounts.
Sources: product, pricing, Zipher on AWS Marketplace
Altimate AI
Per account
Auto Tune resizes job clusters, SQL warehouses and all-purpose compute itself, between runs, and rolls a change back automatically if a run fails or slows down too much. It also splits spend by team and lists other savings with an owner for each. It covers Snowflake too.
Access: A service principal with read-only metadata access. Write access only on the jobs, warehouses and clusters you turn Auto Tune on for, and it changes configuration only, never code, schedules or data. Unity Catalog is required.
Price: Not public. Priced per account with no per-seat fee; a 30-day proof of value is offered.
Sources: Databricks, pricing
Sedai
Per vCPU managed, yearly
Resizes Databricks compute and shutdown settings based on real CPU and memory use, as part of a wider cloud optimization platform, and can roll a change back if performance drops. It does not change queries or code.
Access: Connects through your cloud account (on Azure, an app registration or Sedai’s own service principal, with a reader role). You choose: recommendations only, approve each change, or fully automatic.
Price: Not public. Quoted on how much it manages (it bills per vCPU managed), with a fixed yearly commitment.
Sources: data platforms, pricing, docs: connect Azure
Revefi
Percentage of spend
An AI agent (“AI DBA”) that tracks Databricks spend, tunes jobs through test runs, and resizes clusters and SQL warehouses itself, asking for approval when needed. It also does data quality monitoring and covers Snowflake and BigQuery. Suits smaller teams that fit the free tier, and teams that want cost and data quality monitoring in one tool.
Access: A service principal that reads the system catalog through a SQL warehouse, plus a catalog of its own for cached views. Data quality checks read the catalogs you grant it. Job tuning needs manage permission on jobs and clusters, and cluster logs written to a volume.
Price: Free (Starter) while your combined yearly spend is under $50K. Above that, priced as a percentage of spend; the rate is not public.
Sources: Databricks, pricing, docs: Databricks setup, AI DBA for Databricks (press release)
Capital One SlingshotCapital One Software
Not public
Databricks cost views (by workspace, job and compute), best-practice recommendations, and machine-learning tuning of job clusters against a runtime target, with auto-apply switched on or off per job. The Databricks features came from Sync Computing, whose team joined Capital One Software in 2025. Its product page focuses on Snowflake; the Databricks features are in its docs.
Access: Slingshot says it only reads metadata. Job tuning also reads cluster events from your cloud account (AWS or Azure), and its savings figures use cloud list prices rather than your discounts. The exact Databricks permissions are not in its public docs.
Price: Not public. Book a call through Capital One Software.
Sources: docs: job cluster optimization, docs: project settings (auto-apply), docs: FAQs, Slingshot, Databricks partnership and Sync Computing
SELECTSELECT (part of DoiT)
Custom quote
Cost breakdowns, budgets and spend alerts for Databricks, plus automatic adjustments to jobs compute, all-purpose compute and serverless SQL warehouses. Best known for Snowflake; Databricks support launched in June 2026. DoiT bought SELECT in January 2026. Suits teams on both Snowflake and Databricks that want one cost tool for both.
Access: Setup uses a read-only service principal and takes about 20 minutes, per its launch announcement. The pages do not say what extra access the automatic changes need.
Price: Custom quote, starting at $1,499 a month.
Sources: Databricks, pricing, DoiT launches SELECT for Databricks (press release), DoiT acquires SELECT
Seemore Data
Not public
Cost attribution, alerts on unusual spend and data lineage for Databricks, with AI agents that size, scale and stop SQL warehouses within policies you approve. It started on Snowflake; its Databricks product launched in October 2026 and its page now says it is live.
Access: A service principal you create reads the system tables (metadata only; Seemore says it never reads table data). Write access is optional and limited to starting, stopping and resizing warehouses you approve.
Price: Not public. Free trial and demo offered.
Sources: Databricks, free trial, Seemore Data launches Databricks support (press release)
DoiT PerfectScale for Databricks
Not public
Part of DoiT’s cloud cost platform: cost by job, pipeline, cluster and warehouse, team budgets, spend alerts, and compute that adjusts itself. DoiT owns SELECT, and the two Databricks pages list very similar features. Suits companies already using DoiT for their cloud costs.
Access: Read-only access to usage metadata; DoiT says most teams are live in under 15 minutes.
Price: Not public for Databricks. Free trial offered.
Sources: PerfectScale for Databricks, PerfectScale for data platforms
Usage.ai
Share of savings
Buys and manages Azure reservations and savings plans in your account, including for the virtual machines under classic Azure Databricks compute, and pays back eligible unused commitment value under its programme terms. It does not tune Databricks itself. Suits Azure Databricks teams with steady classic compute that want machine commitments bought and managed for them.
Access: Read-only access to billing and usage data, plus permission to buy commitments once you turn purchasing on. You approve each purchase (Copilot) or let it buy within your limits (Autopilot). It cannot start, stop or resize anything.
Price: Not public. A share of the new savings it generates (the rate is not on its pages), with no upfront fee or subscription and no charge on commitments you already had.
Sources: Azure Databricks commitments, Azure, home and pricing
How to choose
Do you only need to see the spend?
Start with Databricks’ own tools. System tables, usage dashboards, budgets, tags and compute policies come with your account and cover visibility, alerts and limits. For more detail at a published price, look at Lakesight, Vantage and LakeSentry.
Is Databricks one part of a bigger cloud bill?
CloudZero, Vantage, Flexera, PointFive, Sedai and DoiT PerfectScale put Databricks next to your other cloud costs. Usage.ai buys Azure commitments that also cover the machines under classic Databricks compute.
Do you want a tool to change settings for you?
Keebo, Espresso AI, Zipher, Sedai, Seemore Data, SELECT, DoiT PerfectScale, Capital One Slingshot, Altimate AI and Revefi all make changes themselves, and Unravel, PointFive and Lakemine offer automated actions too. Ask what write access each needs, how changes are approved, and how they are undone.
Would you rather your team makes every change?
LakeSentry, Flexera and ClusterBite suggest fixes and leave every change to you, and Lakemine only suggests unless you allow it to act. Sedai can run in a recommendations-only mode, and PointFive, Slingshot and Usage.ai can ask before each change.
How do you want to pay?
Monthly plans at a published price are easy to budget: Lakesight, LakeSentry, and Vantage, whose tier depends on the cloud spend you track. A share of savings ties the cost to what is saved: ClusterBite, Keebo, Espresso AI, Usage.ai, and Lakemine’s pay-as-you-save option. Pricing based on spend or usage depends on how much the tool covers: Unravel, Zipher, Sedai and Revefi. The rest are quoted after a call or don’t publish prices: CloudZero, Flexera, PointFive, Altimate AI, Capital One Slingshot, SELECT (from $1,499 a month), Seemore Data and DoiT PerfectScale.
Can your cost data leave your network?
If not, look for a tool that runs on your side: Lakemine runs inside your Databricks account, LakeSentry offers a Databricks App and a private cloud install, and ClusterBite offers a private deployment inside your network.
Which cloud are you on?
Check each vendor’s supported clouds before you start. ClusterBite supports Azure Databricks today.
Where ClusterBite fits
- Our fee is 30% of what each change saves for the first three months after you make it. After that the whole saving is yours. If a change saves nothing, you pay nothing for it. The pricing page lists a few exceptions, such as a recommendation you never answer.
- The app only has read access and never changes anything in your workspace. Your team makes each change.
- We host it, or we install it inside your network if your cost data has to stay there.
- It runs alongside Databricks’ own tools and the rest of your cost work.
What the fee adds up to
Not like for like: Espresso AI tunes your SQL warehouses itself, while with ClusterBite your team makes each change.
One change that saves $1,000 a month. Fee in the first year:
- ClusterBite (30% × three months, then nothing)
- $900
- Espresso AI, monthly (40% × 12 months, and it carries on)
- $4,800
- Espresso AI, upfront (36% of estimated yearly savings, paid upfront)
- $4,320
Espresso AI is the only other tool here that publishes its savings rate. Its monthly plan bills 40% of each month’s savings, and its pricing page gives no end date. It also offers 36% of estimated yearly savings, paid upfront. Keebo, Usage.ai and Lakemine also charge from savings but don’t publish a rate.