# StackSpend — Machine Version

> StackSpend is an AI Spend Intelligence platform: cost management for the
> modern AI engineering stack — LLM inference, AI coding tools, and the cloud
> they run on. It connects read-only, unifies spend into one daily signal, and
> adds the analysis a FinOps hire would do — anomaly detection, forecasting,
> attribution, a conversational cost analyst, and model-swap recommendations.
> Built for 20–200 person engineering-led companies that have outgrown informal
> oversight but do not have a dedicated FinOps function.

AI Spend Intelligence is the category of cost tooling that goes past dashboards:
it normalises spend across every cloud and AI provider, then applies analysis and
automation on top, so teams get answers and actions rather than charts. The
operating principle is cost accountability for engineering: cost owned at the
point of decision — by the team that created it, the day they created it — not
reported to one person after the money is gone. Self-serve, no sales call.

Last updated: 2026-08-11

---

## What problem it solves

Cloud and AI spend moves with usage, not headcount. A prompt change, a retry
loop, or a model upgrade can move the bill in days — and provider dashboards
show one provider at a time, a day or more behind, with no attribution to your
services or teams.

StackSpend answers four questions daily, across every provider at once:

1. What did we spend yesterday, and is that normal for us?
2. Where is this month landing, against budget?
3. What changed, and which commit or feature caused it?
4. What could we switch to that costs less and scores as well?

---

## Who it is for

The core buyer is a **20–200 person, engineering-led company** where cloud and
AI spend has grown past the point where informal oversight works, but has not
grown enough to justify a dedicated FinOps hire. StackSpend is built to serve
the engineering leader and the finance partner in the same conversation, because
at that size they usually are in the same conversation.

Typical roles that adopt it:

* **CTO / VP Engineering / Engineering Manager** (Series A–C) — no single view of
  spend, reacting to spikes after the invoice.
* **CFO / Finance Manager** (Series B+) — spend that cannot be mapped to teams or
  products, no working budget or forecast.
* **Platform / DevOps / SRE** (50+ people) — investigation is slow and there is no
  way to get cost data into existing tooling.
* **Head of Ops / IT Lead** (30–200 people) — owns billing admin, audit, and
  compliance questions, and gets the surprises.
* **Technical founder** (pre–Series A) — no visibility, AI spend growing fastest.

It is deliberately not built for a single developer tracking a personal project,
nor for a large enterprise estate that needs a full FinOps governance suite. The
shape it fits is several providers, several teams, real money, and nobody whose
full-time job is watching it.

---

## Provider coverage

Connected read-only via API keys, billing exports, OAuth, or OpenTelemetry.
17 native sources, plus anything else through FOCUS import.

**AI:** OpenAI, Anthropic, Fireworks AI, Claude (estimated — see below), Cursor, Hugging Face, Grok (xAI)
**Developer:** GitHub
**Cloud:** AWS, GCP, Azure, Vercel
**Data:** Snowflake, ClickHouse Cloud, Elastic Cloud, Databricks
**Communications:** Twilio

**Billed vs estimated.** Most providers report actual billed cost from their
billing API or export. Agent tools that publish usage telemetry but no
per-seat billing feed — Claude Code and similar — are *estimated*: StackSpend
prices the reported token usage at published API rates. That is a usage-value
figure, not an invoice line, and StackSpend labels it as such rather than
blending it silently into billed spend.

**Anything else:** push spend in the FinOps FOCUS schema via the public API or
a CSV import — custom providers appear alongside native ones everywhere.

**Delivery:** Slack, Microsoft Teams, email.
**Work tracking:** Linear, Jira — an anomaly can open a ticket
with the spend context attached.

---

## Core capabilities

### Daily signal in Slack

* **Problem it solves:** A dashboard only helps the person who remembers to open it.
* One message each morning in Slack, Teams or email: yesterday's spend, budget pace, anything unusual. Green means on track — most days nobody has to think about cost at all, which is the point.
* **How it works:** A daily digest plus anomaly and budget alerts to Slack, Microsoft Teams or email, with per-user opt-in and an org-configurable report time.
* Available on: Team and above.
* Detail: https://www.stackspend.app/cloud-cost-alerts

### Anomaly detection

* **Problem it solves:** Native dashboards don't push. A spike found at invoice time has already been running for weeks.
* StackSpend learns what normal looks like per provider, account and service, then flags the day something breaks pattern, with a severity and an owner. Each one carries a lifecycle, so it gets closed.
* **How it works:** Statistical baselines — mean, standard deviation, median and p95 per provider, account and service — with z-score deviation, a severity, and a lifecycle so it gets closed.
* Available on: All plans.
* Detail: https://www.stackspend.app/spend-anomaly-detection

### Cost Intelligence Agent

* **Problem it solves:** Understanding a bill means half an hour of dashboard archaeology, and there is no FinOps analyst to hand it to.
* Ask in plain English what drove the bill up, or whether you are on budget. The agent answers with cited figures from your own data, and acts only once you confirm.
* **How it works:** Ask in plain English and it chains the queries, answers with exact figures and dates, and takes write actions — acknowledging an anomaly, creating a budget — only after you confirm.
* Available on: All plans.
* Detail: https://www.stackspend.app/ai-explorer

### Forecasting

* **Problem it solves:** “Are we on track?” is a standing board question, and a straight-line guess is not a defensible answer.
* A time-series model projects where the month lands, with confidence bands and the date you would breach budget at current pace. Finance can plan against it while there is still month left to act.
* **How it works:** A Prophet time-series model with upper and lower confidence bounds and a days-to-risk date, per provider and org-wide, falling back to a simple average when data is sparse.
* Available on: Team and above.
* Detail: https://www.stackspend.app/cloud-cost-forecasting

### Budgets and alerts

* **Problem it solves:** Spend grows without thresholds or owners — and a team that doesn't know what normal looks like can't set a budget at all.
* Set budgets at any scope — org-wide, per provider, account, project or tag — and StackSpend watches them daily with alerts at 50, 80 and 100%. If you don't know where to start, auto-budgets seed the numbers from your own history.
* **How it works:** Budgets scoped org-wide, per provider, account, project or tag, with thresholds at 50/80/100% — and auto-budgets that seed the number from your own history.
* Available on: Team and above.
* Detail: https://www.stackspend.app/forecast-budget-track

### Multi-account consolidation

* **Problem it solves:** A serious estate is four Snowflake accounts, three Azure subscriptions and two Anthropic orgs — each with its own console and invoice, and no combined number.
* Connect every account separately and StackSpend does the aggregation you do by hand today: one figure for the estate, broken down by provider, account, service and project. Ask what the company spent yesterday and answer it without opening a console.
* **How it works:** Four Snowflake accounts, three Azure subscriptions, two Anthropic orgs — connected separately, deduplicated on provider line-item ID, and rolled up into a single figure.
* Available on: All plans.
* Detail: https://www.stackspend.app/cloud-cost-monitoring

### Tagging and attribution

* **Problem it solves:** Engineering is one line item. Nobody can say which team, product or customer caused which spend — so nobody owns it.
* Auto-tagging rules label costs as they are ingested, matching provider, account, service and project patterns in priority order. By the time someone asks who owns the spend, the answer is already on the data — filterable and groupable in the explorer.
* **How it works:** Auto-tagging rules match provider, account, service and project patterns in priority order at ingest — so attribution is already there when the question arrives.
* Available on: Team and above.
* Detail: https://www.stackspend.app/model-cost-attribution

### AI Explorer

* **Problem it solves:** Model usage sprawls across providers and coding tools. “Which models are we actually using, and what is that worth?” has no answer.
* Usage by base model, project and user, in tokens and in API-equivalent value. Estimated and billed usage stay separate, so the numbers never double-count and never pretend to be your invoice.
* **How it works:** Usage by provider, base model, project and user in tokens and API-equivalent value, with input/output/cache split and a blended effective rate — estimated and billed kept separate so nothing double-counts.
* Available on: **Business and Enterprise only**.
* Detail: https://www.stackspend.app/ai-explorer

### Model recommendations

* **Problem it solves:** A cheaper model that is just as good for your workload ships every month, and nobody has time to re-evaluate — early picks quietly become a tax.
* StackSpend checks the models you run against a priced, benchmarked catalogue every day. When a cheaper one scores as well, you get the swap, the evidence, and the monthly saving at your real token mix.
* **How it works:** Every model you use is checked daily against the priced and benchmarked catalogue; a candidate must match or beat your model’s strongest axis and stay in tolerance on the rest, priced at your real 30-day token mix.
* Available on: **Business and Enterprise only**.
* Detail: https://www.stackspend.app/model-recommendations

### Jira and Linear

* **Problem it solves:** Cost work that lives in a cost tool doesn't get done.
* Anomalies become Linear or Jira issues assigned to the owner, with severity as priority. Status syncs both ways, and closing the ticket resolves the anomaly.
* **How it works:** Anomalies become assigned issues, status and assignee sync both ways in real time, and closing the ticket resolves the anomaly — with an echo guard so synced changes never bounce back.
* Available on: **Business and Enterprise only**.
* Detail: https://www.stackspend.app/engineering-spend-management

### Audit trail and GDPR

* **Problem it solves:** Change history and erasure requests are compliance requirements, and they land on whoever administers the tool.
* Every meaningful change — auth, providers, team, billing — is logged to a filterable, exportable trail. GDPR export and deletion are self-service, and both are themselves logged. Built for the buyer whose security reviewer actually reads the docs.
* **How it works:** Auth, provider, team and billing events logged to a filterable, exportable trail, with GDPR export and admin-gated deletion — both themselves logged.
* Available on: **Business and Enterprise only**.
* Detail: https://www.stackspend.app/technology-spend-management

---

## The model selection problem

Picking the right model is the single largest cost lever in an AI product,
and it is almost always decided once — early, under time pressure, on the
model that was best that month — then never revisited. Four things make it
hard to revisit:

1. **Price and quality both move monthly.** New models ship, prices are cut,
   and the frontier reshuffles. A choice made two quarters ago is being
   priced against a market that no longer exists.
2. **Headline price per million tokens is the wrong unit.** What you actually
   pay depends on your input/output ratio, how much of your context is
   cache-hit, and whether the cheaper model needs more retries or longer
   outputs to do the same job. Two models with identical rate cards can
   differ severalfold on the same workload.
3. **"Cheaper" and "good enough" are separate questions.** Cost data lives
   with the provider; quality data lives in third-party benchmarks. Nobody
   joins them, so the trade-off gets made on instinct.
4. **A swap is only worth it if it holds.** Savings evaporate quietly when a
   prompt grows, a retry loop appears, or the provider changes its pricing.

How StackSpend addresses it:

* **Price your own mix, not the rate card.** Candidate models are costed
  against your actual token profile — input, output, and cache — so the
  projected saving reflects your workload rather than a list price.
* **Gate on the axis that matters.** Recommendations are checked against the
  benchmark axis relevant to the workload rather than an average score, so a
  model is not proposed for coding on the strength of its maths result.
* **Third-party benchmarks, named.** Coding = best of Aider polyglot and
  SWE-bench Verified; reasoning = GPQA Diamond; maths = best of MATH Level 5
  and Mock AIME (Epoch AI). Prices track LiteLLM. StackSpend does not publish
  its own model rankings — the point is that the numbers are auditable.
* **Re-checked daily,** because the inputs change under you.

**Free and public, no signup.** The underlying price and benchmark data is
published as a reference anyone can use or cite:

* Per-model API prices, refreshed daily: https://www.stackspend.app/resources/llm-api-pricing
* Cost calculator for a given workload: https://www.stackspend.app/resources/llm-cost-calculator
* What changed each month — new models, price moves: https://www.stackspend.app/resources/model-changes
* Model reference and definitions: https://www.stackspend.app/resources/model-glossary

The in-product recommender is what adds *your* usage to that data. The data
itself is open.

---

## Pricing

Most teams land on Team or Business — Starter is a single-seat entry point, not
the shape the product is designed around.

| Plan | Price / user / month | Providers | Seats | Linked accounts | History |
|---|---|---|---|---|---|
| Starter | $29 ($23 billed annually) | 2 | 1 | 3 | 3 months |
| Team | $79 ($63 billed annually) | 5 | 5 | 10 | 12 months |
| Business | $199 ($159 billed annually) | 15 | 15 | 25 | 24 months |
| Enterprise | Custom, invoiced annually | Unlimited | Unlimited | Unlimited | 36 months |

* Free 14-day trial on any plan. No credit card. Doubles as a free cost
  health audit of your stack.
* Team and Business include an account allowance; beyond it, linked accounts
  are $10 each per month, flat.
* API access: **Business and Enterprise only**. Audit logs: **Business and Enterprise only**.
  CSV export: Team and above. Custom tags: Team and above.

---

## Typical setup

1. **Connect** (~5 minutes per provider): read-only API key, billing export,
   or OAuth. No write scopes are requested.
2. **Backfill:** up to 90 days of history pulled on connect, so baselines and
   forecasts work on day one rather than after a month of watching.
3. **Attribute:** spend is classified by provider, service, model, and account
   automatically; add custom tags for team or feature attribution.
4. **Set budgets** per provider or in total.
5. **Route delivery** to a Slack channel, Microsoft Teams, or email.
6. **Act:** anomalies arrive with context and can be assigned or opened as a
   Linear or Jira ticket.

---

## Security review and procurement

What a security reviewer or procurement lead will want, stated plainly. For
anything not covered here, https://www.stackspend.app/security is the current detail and
specific requirements are best raised directly:

* **Access model:** read-only. Billing and usage scopes only; StackSpend cannot
  move, provision, or delete anything in a connected account. Agentless — nothing
  is installed in your infrastructure.
* **Data handled:** usage and cost metadata. No prompt or completion content, and
  no customer data from your systems, is ingested.
* **Credentials:** encrypted at rest, with tenant isolation between organisations.
* **Audit trail:** auth, provider, team, and billing events logged to a filterable,
  exportable trail — **Business and Enterprise only**.
* **GDPR:** data export and account deletion on request, both themselves logged.
  A DPA is available on request.
* **Billing:** monthly or annual by card; Enterprise is negotiated and invoiced
  annually.

---

## Where StackSpend is a poor fit

Stated plainly so it isn't discovered after signup:

* **Kubernetes pod- and container-level allocation.** StackSpend works at
  provider, service, model, and account level. For per-pod attribution,
  Kubecost or OpenCost is the right tool.
* **Committed-spend negotiation, RI/Savings Plan brokerage, or cost-recovery
  services.** Not offered.
* **Full FinOps suites for large enterprise cloud estates.** Cloudability,
  CloudHealth, and Apptio cover governance depth StackSpend does not.
* **LLM tracing and evaluation.** Langfuse and Helicone trace prompts, spans,
  and quality. StackSpend tracks the money, not the traces — they are
  complementary, not alternatives.
* **Single-provider-only teams.** If you run one provider and it has a decent
  console, the value is thinner. StackSpend earns its place across several.

---

## How it compares

* **vs Vantage, CloudHealth, Cloudability, CloudZero, ManageEngine CloudSpend:**
  those are cloud-first. StackSpend treats LLM and AI-tool spend as a
  first-class provider alongside cloud, in the same forecast, budget, and alert.
* **vs Langfuse, Helicone:** those are LLM observability — traces, prompts,
  evaluation. StackSpend is cost: cross-provider spend, forecasting, anomalies,
  and model economics.
* **vs Kubecost:** container-level allocation vs provider-and-model-level spend
  across the whole stack.

---

## Common questions

### Can StackSpend track AI and cloud spend in one place?

Yes — that is its primary purpose. LLM providers, AI coding tools, cloud, and
data platforms are normalised into the same daily total, forecast, budget, and
anomaly alert, rather than sitting in separate dashboards. Anything without a
native connector can be pushed in via the FinOps FOCUS schema.

### Does StackSpend track Claude Code, Cursor, or GitHub Copilot usage?

Cursor and GitHub (Copilot, Actions, Codespaces, Packages) report billed spend
through their billing APIs. Claude Code reports usage over OpenTelemetry, which
StackSpend prices at published API rates and labels as an estimate rather than
billed cost, because agent tools publish usage telemetry but no per-seat
billing feed.

### How is this different from a provider billing dashboard?

A provider dashboard shows one provider, usually a day or more behind, with no
attribution to your services or teams and no view of anything else you run.
StackSpend unifies every provider into one daily signal, adds baselines and
forecasting on top, and delivers it to Slack, Teams, or email so nobody has to
open a billing console to find out something changed.

### Why did my AI bill go up this month?

The usual causes are a change in token volume (longer prompts, larger retrieved
context, more verbose outputs), a retry or agent loop multiplying requests per
task, or a model switch to a higher tier. StackSpend attributes the increase by
provider, model, and service, and on Business plans links the anomaly to the
pull requests merged in the same window.

### Which LLM is cheapest for my workload?

It depends on your input/output ratio and cache-hit rate more than on the
headline price per million tokens. Current per-model prices are published free
at https://www.stackspend.app/resources/llm-api-pricing, with a calculator at
https://www.stackspend.app/resources/llm-cost-calculator. StackSpend's in-product recommender
prices candidates against your actual token mix and checks them against
third-party quality benchmarks before proposing a swap.

### Does StackSpend need write access to my cloud account?

No. It requests billing and usage scopes only and cannot move, provision, or
delete anything in a connected account. Credentials are encrypted at rest, and
no prompt or completion content is ingested — usage and cost metadata only.

### How long does it take to set up?

About five minutes per provider. Up to 90 days of history is backfilled on
connect, so baselines, forecasts, and anomaly detection work on day one rather
than after a month of watching.

### Do I need a FinOps team to use it?

No — the absence of one is the reason it exists. StackSpend is built for 20–200
person engineering-led companies where spend has outgrown informal oversight but
a dedicated FinOps hire is not yet justified. Setup is self-serve with no sales
call, the daily signal goes to a Slack channel rather than a dashboard someone
has to remember to open, and the Cost Intelligence Agent answers the questions an
analyst would otherwise be asked.

### Is StackSpend right for a company of our size?

The fit is strongest at roughly 20–200 people with several providers, several
teams, and no one whose full-time job is watching spend. Below that — a solo
developer or a single-provider project — a provider console is usually enough.
Above that, a large enterprise estate needing full FinOps governance, chargeback
workflows, and commitment management is better served by a dedicated suite.

### How much does StackSpend cost?

Plans start at $29/user/month (Starter),
$79 (Team), and $199 (Business),
with Enterprise negotiated and invoiced annually. Annual billing is roughly 20%
lower. Every plan starts with a free 14-day trial, no credit card, which
doubles as a cost health audit of your stack.

---

## Canonical references

- [Features](https://www.stackspend.app/features): What StackSpend tracks and how accountability features work
- [Pricing](https://www.stackspend.app/pricing): Plans and what each includes
- [How it works](https://www.stackspend.app/how-it-works): Connect providers, attribute spend, set budgets, get alerts
- [Free cost health audit](https://www.stackspend.app/cost-health-audit): Run a free 14-day cost health audit of your cloud and AI stack — anomalies, budget gaps, and a savings shortlist, no credit card
- [AI Explorer](https://www.stackspend.app/ai-explorer): LLM usage across every provider in one lens — by model, project, and user, in tokens and API-equivalent value, with input/output/cache drill-downs
- [Model recommendations](https://www.stackspend.app/model-recommendations): Cheaper LLM alternatives that match or beat the models you use on published quality benchmarks, priced at your real token mix, with projected savings
- [Security](https://www.stackspend.app/security): Security and data-handling practices
- [FAQ](https://www.stackspend.app/faq): Common questions
- [Docs](https://www.stackspend.app/docs): Setup guides and API reference
- [Custom providers (FOCUS)](https://www.stackspend.app/integrations/custom-providers): Track any cost source — push spend in the FinOps FOCUS schema via the public API or a CSV import
- [Authors](https://www.stackspend.app/authors): Who writes StackSpend's cloud and AI cost coverage
- [Full corpus](https://www.stackspend.app/llms-full.txt): Every guide and article in full, for deeper retrieval

## Solutions by category

- [Cloud Cost Monitoring](https://www.stackspend.app/cloud-cost-monitoring): Cloud cost monitoring across AWS, GCP, Azure, Snowflake, Vercel, and ClickHouse Cloud — one daily signal, no separate billing portals.
- [AI / LLM Cost Monitoring](https://www.stackspend.app/ai-cost-monitoring): Unified AI and LLM spend visibility. Before the invoice, not after.
- [Cost Observability](https://www.stackspend.app/cost-observability): Cost observability and analytics for cloud and AI spend — see, attribute, and explain every dollar in one platform.
- [Cost Monitoring API & Webhooks](https://www.stackspend.app/cost-monitoring-api-webhooks): REST API for line items, rollups, and anomalies. Webhooks push anomaly alerts to your systems. Integrate cloud and AI cost data into dashboards, ticketing, or automation.
- [LLM Cost Monitoring](https://www.stackspend.app/llm-cost-monitoring): Monitor LLM spend across OpenAI, Anthropic, Claude, Cursor, Hugging Face, and Grok in one daily signal — before token costs surprise you.
- [AI Coding Tool Cost Monitoring](https://www.stackspend.app/ai-coding-tool-cost-monitoring): Monitor spend across AI developer tools — Cursor, Claude Code, GitHub Copilot, and Actions — in one engineering cost view.
- [Technology Spend Management](https://www.stackspend.app/technology-spend-management): Bring financial discipline to engineering: one view of cloud and AI spend, budgets, forecasting, and daily signals for the whole tech stack.
- [Engineering Spend Management](https://www.stackspend.app/engineering-spend-management): Manage engineering spend across cloud, AI, and developer tools in one view — budgets, forecasts, and daily signals for software teams.
- [Developer Tool Spend Management](https://www.stackspend.app/developer-tool-spend-management): Manage developer-tool spend — Cursor, GitHub Copilot, Actions, Codespaces, Claude Code — in one view with seat tracking and anomaly alerts.
- [Usage-Based SaaS Spend Control](https://www.stackspend.app/usage-based-saas-spend-control): Control every usage-based vendor — tokens, API calls, compute, seats, messages, bandwidth — before the invoice arrives.

## Solutions by provider

- [AWS Cost Monitoring](https://www.stackspend.app/aws-cost-monitoring): AWS cost monitoring delivered to Slack daily — without logging into Cost Explorer.
- [GCP Cost Monitoring](https://www.stackspend.app/gcp-cost-monitoring): BigQuery billing export to Slack alerts. No custom queries required.
- [Azure Cost Monitoring](https://www.stackspend.app/azure-cost-monitoring): Azure cost monitoring across every subscription — one daily cost signal.
- [OpenAI Cost Monitoring](https://www.stackspend.app/openai-cost-monitoring): Know which model is driving your OpenAI bill. Daily.
- [Anthropic Cost Monitoring](https://www.stackspend.app/anthropic-cost-monitoring): Claude API costs are hard to predict. StackSpend makes them visible.
- [Claude Cost Monitoring](https://www.stackspend.app/claude-cost-monitoring): Track Claude Code, Cowork, and Office agent usage from OpenTelemetry in one daily view.
- [Cursor Cost Monitoring](https://www.stackspend.app/cursor-cost-monitoring): See which developers are driving Cursor usage. Per-user, daily.
- [GitHub Cost Monitoring](https://www.stackspend.app/github-cost-monitoring): Actions, Copilot, Codespaces, Packages — one daily GitHub billing view.
- [Hugging Face Cost Monitoring](https://www.stackspend.app/huggingface-cost-monitoring): Idle Endpoints and surprise Jobs. Hugging Face costs caught daily.
- [Twilio Cost Monitoring](https://www.stackspend.app/twilio-cost-monitoring): SMS spikes and Verify overuse. Caught the day they happen.
- [Vercel Cost Monitoring](https://www.stackspend.app/vercel-cost-monitoring): Project, service, tax, and credit visibility from Vercel FOCUS billing charges.
- [Snowflake Cost Monitoring](https://www.stackspend.app/snowflake-cost-monitoring): Snowflake spend by account, service, and currency. Delivered daily.
- [ClickHouse Cloud Cost Monitoring](https://www.stackspend.app/clickhouse-cost-monitoring): ClickHouse Credits, warehouses, services, and ClickPipes in one daily view.
- [Databricks Cost Monitoring](https://www.stackspend.app/databricks-cost-monitoring): Databricks DBU spend by workspace, SKU, and product — jobs, SQL, DLT, and model serving. Delivered daily.
- [Elastic Cloud Cost Monitoring](https://www.stackspend.app/elastic-cost-monitoring): Elastic Cloud spend by deployment, capacity, storage, and data transfer — delivered daily.
- [Grok Cost Monitoring](https://www.stackspend.app/grok-cost-monitoring): Grok spend grows silently. Get a daily signal before the invoice.
- [Fireworks AI Cost Monitoring](https://www.stackspend.app/fireworks-cost-monitoring): Fireworks inference spend moves fast. Get a daily signal — by model and by dollar — before the invoice.

## Solutions by use case

- [Claude Cost Analysis](https://www.stackspend.app/claude-code-cost-analysis): Ask cost questions in Claude. Get real answers without switching tabs.
- [Cloud Cost Alerts](https://www.stackspend.app/cloud-cost-alerts): Get cloud cost and budget alerts across AWS, GCP, Azure, Snowflake, Vercel, and ClickHouse Cloud — the day spend spikes.
- [Cloud Cost Forecasting](https://www.stackspend.app/cloud-cost-forecasting): Forecast where cloud spend lands before month-end across AWS, GCP, Azure, Snowflake, Vercel, and ClickHouse Cloud.
- [Model Cost Attribution](https://www.stackspend.app/model-cost-attribution): Attribute AI spend to the model, feature, team, user, and customer driving it — so every dollar of LLM cost has an owner.
- [AI Unit Economics & ROI](https://www.stackspend.app/ai-unit-economics): Know whether AI spend is creating margin, productivity, or waste — track AI ROI and unit economics, not just total cost.
- [AI Margin Protection](https://www.stackspend.app/ai-margin-protection): Protect product margins as AI usage scales — catch AI cost movement before it shows up in the P&L.
- [Spend Anomaly Detection](https://www.stackspend.app/spend-anomaly-detection): Catch abnormal spend movement across cloud, AI, and developer tools before it becomes an invoice problem.
- [Shadow AI Spend](https://www.stackspend.app/shadow-ai-spend): Find and govern unmanaged AI spend scattered across teams, tools, credit cards, and API keys.
- [AI Agent Cost Control](https://www.stackspend.app/ai-agent-cost-control): Control the cost of agents, tool calls, retries, and multi-step workflows — before loops and retries blow the budget.
- [Vendor Pricing & Usage Change Monitoring](https://www.stackspend.app/vendor-pricing-change-monitoring): Know when provider pricing, model routing, usage tiers, or billing behavior changes — and what it costs you.
- [Replace Cloud & AI Cost Spreadsheets](https://www.stackspend.app/replace-cloud-cost-spreadsheets): Replace cloud and AI cost-tracking spreadsheets with daily spend signals, anomaly alerts, and forecasts.
- [Cloud & AI Cost Allocation](https://www.stackspend.app/cloud-cost-allocation): Split cloud and AI spend across the teams, products, and environments that created it, so every number has an owner instead of landing in one engineering line item.
- [Cost Allocation Tagging & Tag Enforcement](https://www.stackspend.app/cost-allocation-tagging): Make attribution automatic instead of archaeological — tag rules that apply at ingest across every provider, and a coverage number that tells you whether your IaC tagging policy is holding.
- [Showback vs Chargeback for Cloud & AI Spend](https://www.stackspend.app/showback-vs-chargeback): Show each team what it spent, or bill it back to their budget — which model fits a 20–200 person company, and what has to be true before chargeback works.
- [Per-Team Cloud & AI Budgets](https://www.stackspend.app/per-team-cloud-budgets): Give every team its own ceiling and its own alerts, so cost ownership is distributed instead of sitting with the one person who watches the dashboard.
- [Cloud & AI Cost Ownership (RACI)](https://www.stackspend.app/cloud-cost-ownership): Give every cost a named owner, so spend questions route to a person instead of into the void — the accountability model behind cost control at 20–200 people.
- [Cloud & AI Cost Review Cadence](https://www.stackspend.app/cloud-cost-review-cadence): A monthly cost review that actually gets used — what to put in it, who owns it, and how to stop it becoming a slide deck nobody acts on.
- [Cloud & AI Cost Export for the Accounting Close](https://www.stackspend.app/cloud-cost-accounting-export): Get cloud and AI spend into the month-end close in the shape finance needs — P&L categories, your reporting currency, and an export that does not need re-cutting each month.
- [Multi-Currency Cloud & AI Cost Reporting](https://www.stackspend.app/multi-currency-cloud-costs): Providers bill in USD; your budget is in GBP or EUR. Convert at historical rates, map spend to P&L categories, and close the month without a spreadsheet of manual FX.
- [Read-Only, Agentless Cost Monitoring](https://www.stackspend.app/read-only-cost-monitoring): What a cost tool actually gets access to — read-only billing credentials, no agent, no workload access — and the answers your security review will ask for.
- [GDPR, Audit & Compliance for Cost Tooling](https://www.stackspend.app/cost-data-compliance): The compliance answers a non-engineer has to give about a spend tool — what data it holds, who can see it, how changes are recorded, and how to export or delete it.
- [Is Our Cloud & AI Spend Normal for Our Size?](https://www.stackspend.app/cloud-spend-benchmarks): How to tell whether your cloud and AI bill is reasonable for a company your size — and why your own baseline answers the question better than an industry average.
- [Free Cloud & AI Cost Tracking for Teams](https://www.stackspend.app/free-cloud-cost-tool): What a team can get for free, what a free trial actually proves, and how to work out whether a paid cost tool pays for itself at your spend level.
- [Cost Monitoring Set Up in Minutes, Not a Sprint](https://www.stackspend.app/fast-cost-monitoring-setup): Read-only credentials, about 5 minutes per provider, 90 days of history backfilled — what setup actually involves and why it does not need a project plan.
- [SaaS & Vendor Renewal Tracking for Engineering Tools](https://www.stackspend.app/saas-renewal-tracking): See what every engineering vendor actually costs before the renewal lands — usage-based overages, per-seat creep, and the trials that quietly became line items.

## Comparisons

- [StackSpend vs Vantage](https://www.stackspend.app/compare/stackspend-vs-vantage): StackSpend vs Vantage: StackSpend is simpler and more focused. Fixed pricing from $29/mo, cloud + AI in one dashboard, daily Slack or email signals. Vantage offers more providers and enterprise features; StackSpend fits teams that want visibility without complexity.
- [StackSpend vs CloudHealth](https://www.stackspend.app/compare/stackspend-vs-cloudhealth): StackSpend vs CloudHealth: StackSpend is self-serve with fixed $29–199/mo pricing and cloud + AI in one dashboard. CloudHealth is enterprise-focused (VMware), spend-based pricing, and cloud-only. StackSpend fits teams that want to start in 5 minutes without sales.
- [StackSpend vs ManageEngine CloudSpend](https://www.stackspend.app/compare/stackspend-vs-manageengine-cloudspend): StackSpend vs ManageEngine CloudSpend: ManageEngine CloudSpend is a cloud-only cost allocation tool (AWS, Azure, GCP) with business units, budgets, and showback. StackSpend covers cloud and AI providers in one dashboard with daily Slack or email signals, anomaly detection, forecasting, and a REST API — fixed pricing, 5-minute setup.
- [StackSpend vs CloudZero](https://www.stackspend.app/compare/stackspend-vs-cloudzero): StackSpend vs CloudZero: StackSpend unifies cloud and AI costs (OpenAI, Anthropic, Cursor, etc.) in one dashboard with daily signals and attributes spend per customer, feature, and team via tagging. CloudZero offers deeper Kubernetes and shared-cost allocation but does not track AI/LLM providers. StackSpend fits teams that need cloud and AI managed together.
- [StackSpend vs Kubecost](https://www.stackspend.app/compare/stackspend-vs-kubecost): StackSpend vs Kubecost: StackSpend gives you full cloud (AWS, GCP, Azure) and AI (OpenAI, Anthropic, Cursor, GitHub) in one dashboard. Kubecost is Kubernetes-only—no broader cloud infra or AI providers. StackSpend fits teams that need total technology spend visibility.
- [StackSpend vs Cloudability](https://www.stackspend.app/compare/stackspend-vs-cloudability): StackSpend vs Cloudability: StackSpend is self-serve with fixed pricing and cloud + AI in one dashboard. Cloudability (IBM) is enterprise-focused, sales-led, and cloud-only. StackSpend fits teams that want visibility in 5 minutes without enterprise procurement.
- [StackSpend vs Langfuse](https://www.stackspend.app/compare/stackspend-vs-langfuse): StackSpend vs Langfuse: StackSpend is a cloud and AI cost management platform with daily spend signals, provider billing integrations, and fixed pricing. Langfuse is an LLM observability platform for traces, prompts, and evaluations. StackSpend fits teams trying to control spend; Langfuse fits teams debugging LLM behavior.
- [StackSpend vs Helicone](https://www.stackspend.app/compare/stackspend-vs-helicone): StackSpend vs Helicone: StackSpend monitors cloud and AI spend from provider billing data with daily Slack or email signals. Helicone is LLM observability and gateway tooling for request traces, latency, prompts, and usage analytics. StackSpend fits spend control; Helicone fits LLM operations.

## Reference

- [Cost glossary](https://www.stackspend.app/resources/cost-glossary): Definitions — AI Spend Intelligence, FinOps, burn rate, showback/chargeback, and more
- [Model glossary](https://www.stackspend.app/resources/model-glossary): LLM model reference
- [LLM API pricing](https://www.stackspend.app/resources/llm-api-pricing): Current per-model API prices, refreshed daily
- [Blog](https://www.stackspend.app/blog): Guides on cloud + AI cost management