Skip to main content

aztree

Where did my Azure money go? aztree reads your Azure costs and draws them as a treemap: big box, big cost. It's a port of awstree by Branko Petric, which took the idea from disktree by Tobi Lütke. Same idea again, pointed at an Azure subscription.

aztree showing a demo Azure bill as a treemap

It's one small Python package with no dependencies, and it produces a single HTML page that works offline.

Quick start

pipx install aztree         # or: pip install aztree
aztree --demo               # try it with fake data, no Azure needed
az login
aztree                      # your current subscription
aztree --all                # every subscription you can see

From a clone, run python -m aztree without installing anything.

The page opens in your browser. It shows the last 30 days, compared with the 30 days before. The last day is the day before yesterday, because Azure is still posting yesterday's usage.

What you get

  • Treemap of spend by service → meter. Press 2 to group by subscription, 3 by region, 4 by resource group → resource and 5 by tag value → service.
  • Tag view: spend by the values of one tag, with untagged spend as its own box and its share of the bill in the header. aztree picks the tag on the most resources; --tag KEY picks another.
  • Color by change (c): red where spend grew, green where it shrank.
  • Selection panel: cost, share of the bill, change vs the previous period, monthly pace and a daily chart. For the whole bill, the last two cells show this month so far and Azure's forecast for the month.
  • Worth a look: one list where news and to-dos take turns. News is costs that grew and one-off spikes (a day far above that meter's usual). To-dos are known money pits, compute left on all week in dev/test resource groups, and resources that bill while doing nothing, which aztree finds with Azure Resource Graph (VMs stopped but not deallocated, unattached disks, unused public IPs, snapshots older than 90 days, App Service plans with no apps, NAT gateways on no subnet). When Advisor has no reservation or savings-plan tips, steady spend a reservation could cut shows up too. The first six show; "show all" opens the rest.
  • Money pits it knows: Log Analytics ingestion, data transfer out, NAT and Firewall processing, public IPs, snapshots, retiring VM series with their retirement dates, Standard and Premium v2 App Service plans, several Front Door profiles, Front Door Premium, large DTU databases, Azure DevOps seats and hosted jobs, private endpoints, provisioned Cosmos DB throughput, 1-minute alert rules, Azure Cache for Redis (retiring) and Extended Security Updates.
  • Biggest drops: what fell since the previous period, including anything that went to zero, so you can see a saving land.
  • Advisor: Azure Advisor's cost recommendations (reservations, savings plans, right-sizing), one per resource and SKU. Reservation and savings-plan tips list the meters they would cover and what those run at a month. Click a tip to jump to its biggest meter, or to the resource it's about.
  • Export for AI: a JSON summary to hand to any AI agent (see below).

Click to select, double-click to zoom in, ⌫ to go back up, / to filter. Add #resource, #subscription, #region or #tag to the page's URL to open that view.

Requirements

  • Python 3.9+ and the Azure CLI, logged in with az login. Without the CLI, put a token for https://management.azure.com/ in AZURE_ACCESS_TOKEN and name the subscriptions with --subscription or --all. A bare token only sees its own tenant; through the CLI, aztree sees every tenant you're logged in to and gets a token for each.
  • The Cost Management Reader role (or Reader) on each subscription. Advisor tips and the Resource Graph checks need Reader; without it you still get the page.
  • Tested on a CSP subscription. On CSP, Azure shows costs at retail prices without your partner's discounts, and reserved usage as $0 even with --metric AmortizedCost, so the totals won't match your partner's invoice. Pay-as-you-go and Visual Studio subscriptions use the same API.
  • EA and MCA billing scopes should work through --scope, but nobody has tried yet. At such a scope the subscription view shows the whole scope as one box, and Advisor is skipped.

Cost: Cost Management queries are free. Azure throttles them per subscription and per tenant, so a run makes about 10–14 requests per subscription and may wait 30–60 seconds when Azure asks it to. aztree saves the data in ~/.aztree/, so reopening the page is instant:

aztree --from

Ask an AI about your bill

aztree --from --export

This writes ~/.aztree/aztree-export.json, a compact summary that includes instructions for the agent. It has totals, breakdowns by service, subscription, region and resource group, and every meter with its change vs the previous period. It also lists the top growers and drops, credits and refunds, the flagged money pits, idle resources, the tag breakdown, this month's forecast and the Advisor tips. Give it to an AI agent and ask "where can I save money?". The Export for AI button in the page (or e) downloads the same file.

Options

Flag What it does
--demo Fake data, no Azure access needed
--subscription ID_OR_NAME Subscription to read; repeat for more. Default: the Azure CLI's current one
--all Every enabled subscription you can see, in every tenant you're logged in to
--scope SCOPE Any Cost Management scope, e.g. /providers/Microsoft.Billing/billingAccounts/ID (untested)
--days N Period length, default 30. It's always compared with the period before it.
--metric M ActualCost (default) or AmortizedCost
--no-advisor Skip Azure Advisor
--tag KEY Tag for the tag view. Default: the tag on the most resources
--no-graph Skip the Resource Graph checks for idle resources
--from [FILE] Reopen saved data without calling Azure (default: the last run)
--export [FILE] Write the AI summary JSON instead of the page (default ~/.aztree/aztree-export.json)
--out FILE Where to write the page (default ~/.aztree/aztree.html)
--no-open Don't open the browser
--verbose Print pages, rows and query units per request
--version Print the version

Good to know

  • ActualCost books a reservation or savings plan purchase as one lump on the day you bought it. --metric AmortizedCost spreads it over the term.
  • Credits and refunds count in the totals, but negative amounts can't be drawn as boxes. The header says how much of the total they are.
  • Subscriptions that bill in different currencies are drawn in USD.
  • If a subscription has so many resources that a daily breakdown takes more than ten pages of results, the resource view still lists every resource but with one total per period and no daily chart.
  • Marketplace charges keep their publisher's meter names and get no special handling.
  • The forecast is Azure's own, for the calendar month, so it doesn't follow --days. It's left out when a subscription has none, or when it comes in a different currency from the page.
  • Runs saved by older versions open without the tag view, the forecast and the idle checks.
  • Runs save to ~/.aztree/; set AZTREE_HOME to move it. Older versions run from a clone saved to out/, and aztree --from out/aztree-data.json still opens those runs.

Checking against the portal

In the Azure portal, open Cost Management → Cost analysis, choose the cost type you ran with (Actual cost by default), daily granularity and the dates in aztree's header. The totals should match to the cent. Azure can still add a little to the last day after aztree reads it, so compare soon after a run.

Privacy

aztree only reads cost data, forecasts, tag names, Advisor recommendations and Resource Graph properties, and never changes anything in your subscriptions. Everything it writes goes to ~/.aztree/ (or AZTREE_HOME), outside any git repo, because it contains your subscription IDs and costs. The page loads nothing from the internet. Share the HTML or export file only with people who should see your bill.

Development

python -m unittest discover -s tests
python -m aztree --demo

The tests need no Azure access. The viewer tests run the page's script in Node and are skipped without it.

Credits

aztree is a port of awstree by Branko Petric (MIT). It keeps awstree's viewer and its JSON contract and replaces the AWS parts. awstree, in turn, is inspired by disktree by Tobi Lütke.

aztree is not affiliated with or endorsed by Microsoft.

License

MIT

Metadata

Release files for aztree 0.4.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for aztree 0.4.0
File Size Uploaded
aztree-0.4.0.tar.gz 70.7 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for aztree 0.4.0
File Interpreter ABI Platform
aztree-0.4.0-py3-none-any.whl Python 3 none any Details

Total release size: 114.9 kB

Release files / aztree-0.4.0.tar.gz

Download URL aztree-0.4.0.tar.gz
Size 70.7 kB
Tags Source
SHA-256 checksum
How to use checksums
dc13348ae4b367a6569a38d1308b656532e5d3c87d0e9a623b8b0b40e70af862
BLAKE2b-256 checksum
How to use checksums
9f2360629b0c82dfd6ae71e42f474ed776384e7a7d502f274389ac8433e33ba2
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Sep 28, 2026.

Transparency log

Release files / aztree-0.4.0-py3-none-any.whl

Download URL aztree-0.4.0-py3-none-any.whl
Size 44.2 kB
Tags Python 3
SHA-256 checksum
How to use checksums
c253cb7d1ccad949b0d1bbd59afb2cb53b8f7c3e731808128f31a5c8adc0ad59
BLAKE2b-256 checksum
How to use checksums
0845679a5382bde17014d1f71b75126d13ecf017ecc234256f4ef047b44071db
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Sep 28, 2026.

Transparency log

Release history Release notifications | RSS feed

0.9.2

2 release files

0.8.3

2 release files

0.8.2

2 release files

0.8.1

2 release files

0.8.0

2 release files

0.7.0

2 release files

0.6.1

2 release files

0.6.0

2 release files

0.5.3

2 release files

0.5.2

2 release files

0.5.1

2 release files

0.5.0

2 release files

This release

0.4.0 This release

2 release files

0.3.0

2 release files

0.2.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page