> ## Documentation Index
> Fetch the complete documentation index at: https://docs.guild.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Spend Explorer
> Track and attribute your team's Anthropic, OpenAI, and Cursor spend at spend.guild.ai.
Spend Explorer at [spend.guild.ai](https://spend.guild.ai) reports what your team spends on AI coding and inference. Connect the platforms you buy from and it imports their usage and billed cost, attributes the spend to the people and keys that incurred it, and breaks it down by platform, model provider, model, and user.
Spend Explorer is its own product with its own sign-in. It reads usage and cost reports from the platforms you connect, and nothing else.
If you already use Guild, the same platform spend appears inside the product at **Insights → Platform**, alongside the usage Guild records from its own runtime. See [Connect spend platforms](/insights/connect-platforms). Spend Explorer covers external platform spend only.
## Signing in
Sign in with your work email. Spend Explorer sends a sign-in link — there is no password — and the link is valid for 15 minutes.
Spend is reported per organization, which is why the form asks for a work address and turns down consumer email domains.
## Connecting a platform
The setup flow walks through four steps: choose a platform, create an admin key on it, paste the key, and watch the first import.
Guild validates the key against the platform as you save it, so the wait is a few seconds. Keys are stored server-side and never shown again. The first import pulls up to 90 days of history and reports its progress; new spend appears within a few minutes.
Add the rest of your platforms whenever you like — the same flow becomes the manager you use afterwards.
### Which key each platform needs
| Platform | Key | Where to create it | Can you scope it down? |
| ------------- | ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------- |
| **Anthropic** | Admin key (`sk-ant-admin01-…`) | Anthropic Console → Settings → Admin keys → Create admin key. Keys are organization-wide, so an organization admin has to create it. | No — console admin keys carry full admin access. |
| **OpenAI** | Admin key (`sk-admin-…`) | OpenAI platform → Settings → Organization → Admin keys → Create admin key. | Yes — set it to **Read only**, or **Restricted** with just the usage and cost scopes. |
| **Cursor** | Admin API key (`key_…`) | Cursor dashboard → API Keys. Team admins only. | No — Cursor admin keys carry full admin access. |
Each console shows the key once, so copy it before you leave the page.
Anthropic and Cursor admin keys cannot be narrowed, so each one grants full admin access to that account. Store them accordingly and rotate any key you think has been exposed. Where a platform does allow narrowing, as OpenAI does, grant the least it will accept.
## Reading the dashboard
The dashboard covers a time window you choose — 1, 7, 30, or 90 days, or a custom range of up to 180 days. Windows are UTC calendar days.
Four figures sit at the top:
| Tile | What it counts |
| --------------- | ----------------------------------------------------------- |
| **Spend** | Billed cost reported by every connected platform. |
| **Tokens** | Total tokens across the models reported for the window. |
| **Developers** | People and API keys the platforms could attribute spend to. |
| **Active keys** | Platform API keys with spend in the window. |
Below that, daily spend, the monthly trend, and a breakdown card with tabs for **Platform**, **Provider**, **Model**, **Users**, and **API keys**.
Platforms sync hourly. **Sync now** forces one, and each connected platform carries its own state — syncing, synced, behind, or stuck — so a stale figure is visible rather than silent.
## How spend is attributed
A model bought through one platform is often served by another, so the breakdown offers two ways to group the same spend.
* **Platform** groups spend by who billed it. Claude usage bought through Cursor counts as Cursor, because Cursor is what invoices you. Use this tab to reconcile against your invoices.
* **Provider** groups spend by the model's actual provider. That same Claude usage counts as Anthropic. Cursor's own models — ids beginning `cursor` or `composer`, plus its `auto` and `default` routing ids — stay under Cursor, and a model Spend Explorer does not recognize falls under **Other**.
Within a platform's bill, cost is split across models and people by each one's share of the priced tokens, using the list price in force on the day the tokens were used. A later price change never restates a past day. Cost a platform bills at the day level, with no usage to attribute it to, is held separately rather than being spread onto an arbitrary model.
## What the reports do not cover
Spend Explorer sees what each platform's admin API reports, which is direct API usage billed to the organization the key belongs to. Seat-based and gateway usage is not in those reports — Claude Code subscriptions, Bedrock, and Vertex do not appear in Anthropic's cost reports, and usage billed to another organization does not appear in any of them.
An import that finishes with no usage usually means one of those, or a brand-new account. Check that the key was created in the organization that carries your spend.
## Removing a platform
Remove a key and no further data is imported from that platform. Spend already synced stays on the dashboard.