AI · Token Economics

AI cost visibility by team, model, service and provider

Glassity reads your AI usage from the tools you already run, reports tokens and cost hourly, and puts it in the same format as your cloud bill. Free tier, read-only, fifteen minutes.

Start free Or book a walkthrough

Read-only access · EU-hosted · We never store prompt content

ISO 27001 Certified AWS Qualified Software Data held in AWS eu-west-1, Ireland

What you get

Glassity reads AI usage from the gateway or cloud AI service you already run, shows it hourly per key, model and provider, and writes it nightly into FOCUS next to your cloud bill, where the Glassity Agent can answer questions about both.

  • Tokens and cost per source, key, model and provider, updated hourly
  • Token counts by class: input, output, cache read, cache creation, reasoning
  • Five gateways and two cloud AI services: LiteLLM, Helicone, OpenRouter, Cloudflare AI Gateway, Bifrost, Amazon Bedrock, Microsoft Foundry
  • A usage preview before you connect anything
  • Read-only credentials, hourly aggregates only, never prompt or response content
  • Self-serve setup, nobody from Glassity needs to be on a call

There’s also managed by Glassity, our own gateway, which shows what the Glassity Agent itself costs you.

What each source carries, field by field

Gateways give per-key detail. Amazon Bedrock and Microsoft Foundry give tokens and cost per account, region and model. This is what Glassity can read from each source:

Source Type What you can group by
LiteLLM Gateway Team id, end user, request tags, spend-log metadata
Helicone Gateway User id, custom properties, prompt id
Bifrost Gateway Custom metadata sent in request headers
Cloudflare AI Gateway Gateway Gateway id, request type
OpenRouter Gateway Key only
Amazon Bedrock Cloud AI service Account, region, model; input and output tokens
Microsoft Foundry Cloud AI service Account, region, model

Every gateway row also carries the key id, key name, model and provider. Token counts are split by class: input, output, cache read, cache creation and reasoning.

The Amazon Bedrock bill shows input and output tokens only, and they are real counts, not estimates. Cache and reasoning tokens exist only in gateway data. With the optional CloudWatch role, Bedrock counts arrive live, within minutes.

No field is shared by every source. LiteLLM has a team id, Helicone has custom properties, OpenRouter has only the key. That’s why key naming matters so much: it’s the one thing that works the same way everywhere. How to attribute AI costs to teams when there is no team field →

How much history you get

Amazon Bedrock usage can be read back up to six months, because it is on your AWS bill. You see half a year of history on the first day. Gateway usage cannot: when you connect a gateway, Glassity picks up the last 48 hours and everything from then on, so it’s worth connecting early. Why your Amazon Bedrock bill has token counts but no owner →

How the FOCUS rows are built

Every hour, Glassity stores aggregates per source, key, model and provider: token counts by class, request counts and cost. Never prompt or response content.

Every night, those hourly aggregates are written into FOCUS, the FinOps Foundation’s open billing format, as standard rows next to your cloud billing data. Your AI costs and your cloud costs end up in one place, in one format.

Because they are standard rows, the Glassity Agent and our open-source FOCUS MCP read your token spend without a second data model.

The AI Usage page updates every hour. The FOCUS data lake is written once a night, so the Agent’s answers can use data up to 24 hours older than what the AI Usage page shows.

Ask why the AI bill went up and get an answer

Type the question the way you’d say it. The Glassity Agent reads both your AI usage and your cloud spend and answers in seconds, instead of someone exporting data and building a report.

  • Why did AI cost go up last week, and which key did it?
  • Which model burned the most reasoning tokens this month?
  • What did checkout spend across cloud and AI in August?
  • What share of our Amazon Bedrock spend went through the gateway last month?
  • Cost per million tokens by provider, this month against last.

Cross-dataset answers depend on your own naming.

What it doesn’t do yet

  • We can show you what your AI costs and who spent it. We can’t yet show you what that spending gave you back.
  • You can’t export the data to your own storage yet. For now, you reach it through the Glassity Agent and the open-source FOCUS MCP.
  • Your gateway needs to be reachable over HTTPS from the internet. Gateways kept inside a private network aren’t supported yet.
  • Direct calls to OpenAI, Anthropic or Google aren’t a source. They only show up if they go through a supported gateway or a cloud AI service.

How to start with Glassity: fifteen minutes, read-only

01
Connect your AWS account
A cross-account IAM role with read permissions. No agents, nothing deployed in your environment.
02
Connect your AI usage
Paste the URL and key for your gateway, or use the AWS account already connected for Amazon Bedrock. You see a preview of your own usage before anything is imported.
03
Look at what comes back
Findings by service, AI usage by key, team and model, and an agent you can ask about either.

New to the topic? Start with What is AI token economics? →

FAQ

AI usage is the part of Glassity that reads how many tokens your AI is using and what they cost, shows it hourly by source, key, model and provider, and writes it into FOCUS format next to your cloud billing data so the Glassity Agent can answer questions about both.

Glassity reads AI usage from five gateways and two cloud AI services: LiteLLM, Helicone, OpenRouter, Cloudflare AI Gateway, Bifrost, Amazon Bedrock and Microsoft Foundry. Each connection has a Verify step, a preview of your usage, and then an hourly import.

AI cost by team can be seen through a gateway. Every row Glassity reads carries key id, key name, model and provider, plus that gateway’s own fields such as a LiteLLM team id or a Helicone user id. Amazon Bedrock and Microsoft Foundry attribute to account, region and model instead. There is no single team field shared across gateways, so naming keys or tags consistently is what makes one view work everywhere.

Yes. Bedrock is metered in tokens on the AWS bill, so Glassity turns those lines into input and output tokens per model, account and region, hourly. With the optional CloudWatch role you also get live counts.

Not yet. Glassity reads AI usage from supported gateways and from the Amazon Bedrock and Microsoft Foundry bills. Calls made straight to a provider’s API appear only if they route through one of those. Route them through a supported gateway and they arrive within the hour.

Read-only credentials, encrypted at rest: a gateway key, a monitoring reader, or a CloudWatch role that can only read metrics. Glassity stores hourly aggregates of tokens, requests and cost. Never prompt or response content. Data is held in AWS eu-west-1, Ireland.

AI usage is already included in Glassity’s free tier. Connect a source, see the preview of your own usage, and pay nothing. No card and no call are required.

Token economics, or tokenomics, is the practice of managing the production, consumption and value of AI spend. The Tokenomics Foundation, a Linux Foundation project, defines it as converting energy and capital into AI capability and consuming that capability to produce measurable business value. Tokens are the unit AI is metered in, and the most visible part of AI cost, though not all of it. Glassity calls the reporting side of this AI usage.

No. In AI, tokenomics means managing the cost and value of AI model usage, measured in the tokens models are billed in. It is unrelated to the cryptocurrency use of the word.

See what's in your AI bill

Free tier. Connect read-only in fifteen minutes. No card, no call, and nothing changes in your environment without your approval.

Start free Or book a walkthrough