Lowtide
Book a call (opens in a new tab)Start free
Burnguard · LLM spend and model alerts

Catch runaway AI bills before the invoice does

Burnguard reads your OpenAI and Anthropic cost reports every morning. It emails you when a day spikes, when the month is heading over budget, and before a model you use is retired.

Start free Watch a run

Needs an organisation admin key. Read-only: it only calls cost endpoints.

Works withOpenAIAnthropicCost reports APIAdmin keys (read-only)
46.9%
of enterprises say their actual AI spend is above their planned AI budget.Source: Futurum CIO & Technology Buyers survey, September 2026 (1,636 respondents) (opens in a new tab)
11 Dec
2026: OpenAI shuts down GPT-5, GPT-5 mini, GPT-5 nano and o3 snapshots. Calls to retired models fail.Source: OpenAI API deprecations page (opens in a new tab)
06:17
every morning Burnguard checks yesterday's spend, so a runaway loop is caught the next day, not at month end.How Burnguard works

How it works

From set-up to the first alert

  1. 01

    Connect an admin key

    Add an OpenAI or Anthropic organisation admin key. Burnguard checks it can read cost reports, then stores it encrypted.

  2. 02

    Set a budget

    Optional: a monthly budget per account. Burnguard projects the month from completed days.

  3. 03

    Daily check

    Every morning it reads 35 days of costs, broken down by model, and compares yesterday with the 7 days before.

  4. 04

    Alerts that matter

    An email for spikes over 2× your average, budget overruns, and models retiring in 90, 30 and 7 days.

Live demo

Watch Burnguard work

A real run, step by step, using sample data for a fictional company.

Burnguard · Daily LLM spend checkRunning
Sample data for a fictional company
OPENAI SPEND · LAST 14 DAYS
$184.204.2× normal
Projected $1,912Budget $1,500

What you get

Built for the job, nothing extra

Spikes

Know the morning after, not at month end

A leaked key or a loop that retries forever can burn a month's budget overnight. Burnguard compares yesterday with your normal and tells you fast.

  • Spike rule: over twice the 7-day average
  • Ignores tiny accounts' noise (under $5)
  • One alert per day, never a flood
Model retirements

Never get caught by a retired model

Providers retire models on fixed dates and requests simply fail. Burnguard sees which models you actually use and warns you 90, 30 and 7 days ahead, with the replacement.

  • Models in use with 30-day spend
  • Dates refreshed weekly from provider pages
  • Suggested replacement for each
Budgets

A month projection you can trust

Burnguard projects from completed days only, so today's partial spend never inflates the forecast or triggers a false alarm.

  • Per-account monthly budget
  • Month-to-date and projection
  • Warning while there's still time to act

Free tool

Which AI models stop working next?

Every upcoming OpenAI, Anthropic and Gemini shutdown in one calendar. Paste your model IDs to check them.

Open the calendar

Who it's for

Made for the people who get the blame

“You find out about a runaway job when finance asks about the invoice.”

  • Next-morning spike alerts
  • Spend by model
  • Retirement warnings before calls fail

Industry references

The standards and research behind it

Futurum: AI spend over budget (Sept 2026) (opens in a new tab)Nearly half of enterprises report AI spend above plan.OpenAI API deprecations (opens in a new tab)The official shutdown schedule Burnguard tracks.Anthropic model deprecations (opens in a new tab)Claude model retirement dates and replacements.Read: why we built Burnguard The AI bill nobody checked until the end of the month

Security and data

Careful by default

Read-only by designBurnguard only calls the cost report endpoints. It never sends prompts, changes settings or touches your data.
Encrypted keysAdmin keys are encrypted at rest (AES-256-GCM) and deleted the moment you disconnect.
Minimal dataWe store daily totals and per-model 30-day spend. Nothing about your prompts or users.

Operated by Cloudly Move Ltd in the UK. See our privacy notice.

FAQ

About Burnguard

What key do I need?

An OpenAI organisation Admin key, or an Anthropic Admin API key (sk-ant-admin…). Both are created in the provider's console.

Which models are covered for retirements?

Every model with a published shutdown date on the OpenAI, Anthropic and Google Gemini deprecation pages, refreshed weekly. See the public calendar at /model-retirements.

Will it slow down my app?

No. Burnguard never sits between your app and the provider; it only reads cost reports once a day.

Does it support Azure OpenAI or Bedrock?

Not yet. Those platforms have their own billing and retirement schedules.

Also from Lowtide

Pair it with another agent

MentionedAI search visibilityLearn more OpentideWebsite accessibilityLearn more MooringRecords conformanceLearn more SoundingsAI due diligenceLearn more

Put a safety net under your AI bill

Free during early access. Two minutes to connect.

Prefer to talk it through? Book a call (opens in a new tab)