Helicone cost monitoring
CostMon reads cost, tokens and model from Helicone's gateway request log. The gateway knows which user, app or prompt a call belongs to. CostMon adds that detail to your daily view. An OpenAI or Anthropic invoice cannot give you it.
An OpenAI or Anthropic invoice shows one org-level monthly total. It has no idea which of your users, features, or customers drove it. A SaaS company can sell an AI feature at a flat per-seat price. It cannot tell if a few customers cause most of the spend until someone builds the split by hand. Helicone sits in front of every call and can capture that split. It needs a custom property on each request that names the owner. Something must also bring that gateway log in with the rest of your stack.
Cost anatomy
Alerts
CostMon's default rule watches every service on your account from the day you connect it. An alert always shows up in your CostMon feed. Add an email, Slack, or webhook channel to send it somewhere else too.
A single customer's calls growing heavier per request while every customer still pays the same flat price.
A runaway feature or agent calling one model far more often than its usual traffic.
A monthly budget on total spend routed through Helicone catches growth across every upstream provider it proxies.
Sample alert, illustrative
Anomaly: helicone / gpt-5 spend spike helicone / gpt-5 was $75.00 on 2026-06-22 — 88% above its 14-day average of $40.00. Severity: warning Occurred on: 2026-06-22
To find what to cut, the recommendations list flags idle spend and commitment gaps to review.
Why CostMon
CostMon reads Helicone's gateway request log. It turns the custom properties you name on the connector into cost dimensions. It adds a workload property by default. You can add more, such as user or customer. Spend splits by what you care about instead of one flat provider total.
A per-user view shows the accounts driving a disproportionate share of AI spend while paying the same flat price as everyone else. The pattern is visible within days of the first heavy week.
Helicone knows which customer made each call. CostMon puts that view beside your direct OpenAI and Anthropic bills. Margin questions and vendor totals sit in one place.
CostMon connects with a read-only Helicone API key scoped to request cost and usage data only. It never sees prompts, completions, or account settings.
Getting started
Add a read-only Helicone API key scoped to request cost and usage data. CostMon never sees prompts, completions, or your Helicone account settings.
CostMon pulls per-request rows with cost, tokens, model and your named custom properties. It rolls them into a daily view by model, upstream provider and workload.
Break gateway spend down by customer or workload each day. When one customer's calls get heavier, tokens per request rise and a rule fires.
FAQ
It pulls Helicone's gateway request log: cost, tokens and model for each request. CostMon groups these by day, model, upstream provider and workload. Any other custom property you name on the connector becomes an extra cost dimension.
Two things. First, tag requests with Helicone custom properties (the Helicone-Property-* headers). Use one for whatever you want to attribute by, such as a customer id, a feature name or a prompt version. Second, name those properties on the CostMon connector. They then become cost dimensions. CostMon carries a workload property by default, plus any extras you add. It drops the properties you do not name. Untagged traffic still syncs and counts in your totals. It has no breakdown.
Helicone's dashboard shows gateway usage on its own. CostMon reads that same data. It puts it next to AWS, Anthropic, OpenAI and your other providers. AI spend that goes through Helicone sits in one daily view with the rest.
No. The connector reads cost and usage data only, scoped to a read-only Helicone API key. It never requests access to prompt or completion content.
A flat monthly rate per plan, never a cut of your AI spend or the savings CostMon helps you find. Free includes 2 connectors; Professional and Enterprise include unlimited connectors.
Connectors
AWS, GCP, and Azure for cloud. Anthropic, OpenAI, Amazon Bedrock, Vertex AI / Gemini, and Helicone for AI and LLM spend. Snowflake and Databricks for the data cloud. Datadog, GitHub, and Vercel for the rest of the stack. Add CSV import for any tool without a native connector. That's 14 connectors, all normalized into the same unified view.
Connect a provider and CostMon builds a daily baseline for it. From there, a spike reaches your team as an alert.