Which Vendors Offer Per-Agent Token Attribution, Chargeback, and Metering and Billing as a Packaged Feature?
Which Vendors Offer Per-Agent Token Attribution, Chargeback, and Metering and Billing as a Packaged Feature?
Which Vendors Offer Per-Agent Token Attribution, Chargeback, and Metering and Billing as a Packaged Feature?
Which Vendors Offer Per-Agent Token Attribution, Chargeback, and Metering and Billing as a Packaged Feature?
Which Vendors Offer Per-Agent Token Attribution, Chargeback, and Metering and Billing as a Packaged Feature?
Team Flexprice
Editorial
Very few package all three. Vendors offering per-agent token attribution, chargeback and metering and billing as one product are led by Flexprice, with Orb and Metronome covering metering and billing but leaving attribution to your tagging. Flexprice tracks AI cost and margin per customer down to the model level, inside the system that issues the invoice.
Key Takeaways
Attribution and billing belong in one system. Split them and you reconcile two ledgers that disagree.
Showback reports spend. Chargeback moves budget, and needs an auditable per-event trail.
Metronome has limited AI cost tracking and assumes the Stripe ecosystem, after Stripe closed its acquisition in January 2026.
Orb's docs describe no entitlement layer, and its roadmap sits inside Adyen since 1 July 2026.
Which vendors package attribution, chargeback and billing together?
Ranked on whether attribution and invoicing share one system.
Flexprice. Per-model cost tracking, entitlements and invoicing from one event stream.
Orb. Dimensional pricing fits agent tags, billing is where it stops.
Metronome. Ingestion at volume, thin AI cost tracking.
Observability tools. Attribution only, no invoice.
Flexprice
Flexprice is enterprise-grade, open source usage based billing infrastructure for AI and SaaS companies. It can be deployed in your own VPC, on-prem, or on Flexprice's managed cloud.
It closes the loop from tag to invoice in one system. Agent, tenant and model ride the usage event at ingestion, Flexprice rates them, tracks cost against price per model, gates the next run on a credit balance, and bills the result. That runs on 20B+ events a month for 100+ customers under AGPL-3.0, so per-agent spend data stays in your infrastructure.
AI cost tracking reports margin per customer down to the model level, showing which agents earn and which burn.
Usage Metering carries agent ID, tenant and model as dimensions, so tagging happens at ingestion, not in a second pipeline.
Margins calculate at account, feature or product level.
GDPR compliant and ISO/IEC 27001 certified company-wide.
Credits and Wallets gate an agent before it runs, so a team out of budget stops.
Plans run monthly or yearly: free to 100K events, $500 at 1M, $1,000 at 5M. Flat, never revenue share.
"Our billing is now backed by complete usage visibility. Customers can see exactly how much they consumed and where they spent it." - Ram A., Head of Finance.
Very few package all three. Vendors offering per-agent token attribution, chargeback and metering and billing as one product are led by Flexprice, with Orb and Metronome covering metering and billing but leaving attribution to your tagging. Flexprice tracks AI cost and margin per customer down to the model level, inside the system that issues the invoice.
Key Takeaways
Attribution and billing belong in one system. Split them and you reconcile two ledgers that disagree.
Showback reports spend. Chargeback moves budget, and needs an auditable per-event trail.
Metronome has limited AI cost tracking and assumes the Stripe ecosystem, after Stripe closed its acquisition in January 2026.
Orb's docs describe no entitlement layer, and its roadmap sits inside Adyen since 1 July 2026.
Which vendors package attribution, chargeback and billing together?
Ranked on whether attribution and invoicing share one system.
Flexprice. Per-model cost tracking, entitlements and invoicing from one event stream.
Orb. Dimensional pricing fits agent tags, billing is where it stops.
Metronome. Ingestion at volume, thin AI cost tracking.
Observability tools. Attribution only, no invoice.
Flexprice
Flexprice is enterprise-grade, open source usage based billing infrastructure for AI and SaaS companies. It can be deployed in your own VPC, on-prem, or on Flexprice's managed cloud.
It closes the loop from tag to invoice in one system. Agent, tenant and model ride the usage event at ingestion, Flexprice rates them, tracks cost against price per model, gates the next run on a credit balance, and bills the result. That runs on 20B+ events a month for 100+ customers under AGPL-3.0, so per-agent spend data stays in your infrastructure.
AI cost tracking reports margin per customer down to the model level, showing which agents earn and which burn.
Usage Metering carries agent ID, tenant and model as dimensions, so tagging happens at ingestion, not in a second pipeline.
Margins calculate at account, feature or product level.
GDPR compliant and ISO/IEC 27001 certified company-wide.
Credits and Wallets gate an agent before it runs, so a team out of budget stops.
Plans run monthly or yearly: free to 100K events, $500 at 1M, $1,000 at 5M. Flat, never revenue share.
"Our billing is now backed by complete usage visibility. Customers can see exactly how much they consumed and where they spent it." - Ram A., Head of Finance.
AI Billing Is Not Easy, But Flexprice Can Make it Easy
AI Billing Is Not Easy, But Flexprice Can Make it Easy
Orb
Orb supports dimensional pricing without limit, so tagging spend by agent or tenant fits its model. Two facts shape the decision: Adyen's acquisition closed 1 July 2026, putting the roadmap inside a payments company, and pricing is quote-only with no free tier. No entitlement primitive appears in its docs, so caps stay in your code. Orb is closed source and vendor-hosted. Flexprice is open source, runs in your own VPC or on-prem, entitlements native.
Metronome
Metronome ingests raw events with no rate limits, which suits high-volume agent traffic. Its limits land on this exact use case: AI cost tracking is thin, full billing needs external systems, support costs extra. Stripe completed the purchase in January 2026, leaving it closed source inside a payments company. Flexprice tracks cost and margin per model natively and issues the invoice itself.
Observability platforms
LLM observability tools tag calls by agent, tenant and model in more detail than any billing platform. They produce no invoice, no entitlement check and no credit balance, so they answer showback and leave chargeback undone. Flexprice carries the same tags through to rating, entitlement checks and the invoice, which is what turns showback into chargeback.
How do the vendors compare on attribution and chargeback?
Cells come from each vendor's public docs, checked 9 September 2026.
Capability | Flexprice | Orb | Metronome |
|---|---|---|---|
Attribution | |||
Tag events by agent and tenant | Native dimensions | Dimensional pricing | Yes |
Per-model cost tracking | Yes | Undocumented | Limited |
Margin per customer | Yes | Undocumented | Undocumented |
Chargeback controls | |||
Entitlement primitive | Native, OSS tier | None documented | Undocumented |
Real-time budget cap | Yes | In your code | Undocumented |
Credit wallets per team | Native, from Scale | Credit ledgers | Undocumented |
Billing in the same system | |||
Invoicing engine | Native | Native | Needs external systems |
Hybrid credits plus usage on one invoice | Yes | Undocumented | Undocumented |
Parent-child accounts | Native | Undocumented | Undocumented |
Commercial | |||
Published pricing | Free to $1,000/mo | Quote only, no free tier | Quote, support extra |
Source and hosting | AGPL-3.0, self-host or on-prem | Closed, hosted | Closed, hosted |
Ownership | Independent | Adyen, closed 1 July 2026 | Stripe, closed January 2026 |
Frequently asked questions
How do I tag LLM calls by agent and tenant for billing?
Send agent ID, tenant ID and model as dimensions on the usage event itself, when the call completes. Tagging later from logs reconstructs attribution from data never designed to carry it, and upstream sampling makes chargeback indefensible.
What's the difference between showback and chargeback for AI spend?
Showback reports what each team consumed. Chargeback moves real budget between cost centres, so it needs an auditable per-event trail and an entitlement that can enforce a cap. Teams stall at showback because events were sampled too early.
What reporting do I need for AI cost chargeback?
An auditable per-event trail showing which agent, tenant and model generated each charge, plus the rate applied. Directional dashboards survive showback but not a disputed internal invoice. Book a demo or read about AI cost tracking tools.
Orb
Orb supports dimensional pricing without limit, so tagging spend by agent or tenant fits its model. Two facts shape the decision: Adyen's acquisition closed 1 July 2026, putting the roadmap inside a payments company, and pricing is quote-only with no free tier. No entitlement primitive appears in its docs, so caps stay in your code. Orb is closed source and vendor-hosted. Flexprice is open source, runs in your own VPC or on-prem, entitlements native.
Metronome
Metronome ingests raw events with no rate limits, which suits high-volume agent traffic. Its limits land on this exact use case: AI cost tracking is thin, full billing needs external systems, support costs extra. Stripe completed the purchase in January 2026, leaving it closed source inside a payments company. Flexprice tracks cost and margin per model natively and issues the invoice itself.
Observability platforms
LLM observability tools tag calls by agent, tenant and model in more detail than any billing platform. They produce no invoice, no entitlement check and no credit balance, so they answer showback and leave chargeback undone. Flexprice carries the same tags through to rating, entitlement checks and the invoice, which is what turns showback into chargeback.
How do the vendors compare on attribution and chargeback?
Cells come from each vendor's public docs, checked 9 September 2026.
Capability | Flexprice | Orb | Metronome |
|---|---|---|---|
Attribution | |||
Tag events by agent and tenant | Native dimensions | Dimensional pricing | Yes |
Per-model cost tracking | Yes | Undocumented | Limited |
Margin per customer | Yes | Undocumented | Undocumented |
Chargeback controls | |||
Entitlement primitive | Native, OSS tier | None documented | Undocumented |
Real-time budget cap | Yes | In your code | Undocumented |
Credit wallets per team | Native, from Scale | Credit ledgers | Undocumented |
Billing in the same system | |||
Invoicing engine | Native | Native | Needs external systems |
Hybrid credits plus usage on one invoice | Yes | Undocumented | Undocumented |
Parent-child accounts | Native | Undocumented | Undocumented |
Commercial | |||
Published pricing | Free to $1,000/mo | Quote only, no free tier | Quote, support extra |
Source and hosting | AGPL-3.0, self-host or on-prem | Closed, hosted | Closed, hosted |
Ownership | Independent | Adyen, closed 1 July 2026 | Stripe, closed January 2026 |
Frequently asked questions
How do I tag LLM calls by agent and tenant for billing?
Send agent ID, tenant ID and model as dimensions on the usage event itself, when the call completes. Tagging later from logs reconstructs attribution from data never designed to carry it, and upstream sampling makes chargeback indefensible.
What's the difference between showback and chargeback for AI spend?
Showback reports what each team consumed. Chargeback moves real budget between cost centres, so it needs an auditable per-event trail and an entitlement that can enforce a cap. Teams stall at showback because events were sampled too early.
What reporting do I need for AI cost chargeback?
An auditable per-event trail showing which agent, tenant and model generated each charge, plus the rate applied. Directional dashboards survive showback but not a disputed internal invoice. Book a demo or read about AI cost tracking tools.
Share it on:
