Tollgate joins what you charge with what every run actually costs — live gross margin per customer, per agent, per run. So a healthy blended number can't hide the one that's bleeding.
Customer profitability · 7d
live0.0%
+2.1 ptsCustomer A
rev $0.50 · cost $0.02/run
Customer B
rev $0.50 · cost $0.53/run
Customer C
rev $0.50 · cost $0.11/run
Customer B is unprofitable. A and B are on the same plan — blended metrics hide the loss.
Customer A earns 96% margin. Customer B loses 6%. Both are on the same plan. Blended metrics say everything is fine. Per-customer profitability tells you the truth.
Customer A earns 96%. Customer B loses 6%. Blended, that reads as a comfortable 45%. By the time you spot B, a quarter of profit is gone.
Stripe shows what you charged. Your LLM provider shows what you spent. Nobody shows the difference — per customer, per agent, per run.
One reasoning-heavy run can cost 30× a normal one. Without per-run accounting, you can’t price — you can only guess.
Manual reconciliation survives five customers. At fifty it quietly produces wrong numbers — and wrong numbers compound.
Features
See which customers generate healthy margins and which ones consume more AI spend than they produce in revenue. Per customer, per agent, per workflow.
Acme Corp
Support Agent
Nimbus Inc
Billing Disputes
Quanta AI
Data Analyzer
Track input, output, reasoning, cached, and tool-call costs automatically across providers. Every run accounted for, in real time.
tollgate.track({
customerId: 'acme-corp',
agentId: 'faq-bot',
model: 'claude-sonnet-4-6',
tokensIn: 3000,
tokensOut: 500,
reasoning: 0,
cached: 0,
cost: 0.0165, // cents-accurate
});Get notified when profitability drops, costs spike, or customers become unprofitable. Before the problem compounds.
Customer unprofitable
Nimbus Inc · Billing Disputes
Cost spike detected
Quanta AI · Data Analyzer
Profitability trend declining
All customers
Connects to OpenAI, Anthropic, Google, and any LLM provider. All cost data in one place, normalized and ready for analysis.
Every AI interaction becomes a data point. Revenue minus cost, computed continuously, broken down to the customer and run level.
Surfaces the customers, agents, and workflows that erode margins — so you can act on data, not intuition.
However you price your AI product, Tollgate computes profitability at the unit level.
Link your AI providers — OpenAI, Anthropic, Google, or any LLM. Tollgate starts collecting cost data automatically.
Set up how you charge each customer. Per-seat, per-ticket, usage-based, or custom. Tollgate matches revenue to cost.
See real-time unit economics by customer, agent, and workflow. Get alerts when profitability drops.
From seed-stage startups to growth-stage companies — if you sell AI, you need to know the economics of every customer.
Founders & CEOs
Board-ready unit economics for pricing and fundraising decisions
CFOs & Finance
Real-time cost accounting without monthly spreadsheet archaeology
Revenue leaders
See which customers to renegotiate, expand, or quietly offboard
Product & Eng
Understand the cost impact of every model and workflow change
Tollgate collects only usage and cost metadata. No prompts stored. No conversation content stored. Only token counts and cost data ever leave your stack — never the content of a single request. Delete everything at any time.
Only token counts, model identifiers, and cost data. No prompts, no conversation content, no PII.
Data encrypted in transit and at rest. Access is logged, limited, and auditable.
Delete your data at any time. Disconnect providers instantly. We remove everything from active systems within 48 hours.
Connect your AI stack and get your first profitability report in minutes.
Free during beta · No credit card required