The metering is accurate, yet exporting usage to a billing system remains optional rather than built-in.
Questions about FluxMeter
What is FluxMeter?
FluxMeter is a information technology and services listed on Trustburn. FluxMeter is an open-source usage metering tool designed for AI gateways and multi-tenant products. It tracks usage at the customer and model level, attributing token consumption and cost without providing a full billing suite.
How do I get started with the supported stack?
The supported full stack for metering runs locally with docker-compose via the make demo command. This stack includes HTTP Custody, Kafka, Flink, and Redis, with ClickHouse serving as an audit cold store rather than the metering source of truth. The repository is hosted on GitHub and can be cloned and executed without needing a provider key.
Who is the primary target audience for FluxMeter?
FluxMeter is specifically built for AI gateways and token resellers where one upstream AI account serves many downstream customers. It solves the problem of provider invoices being account-level by tagging every token event to the correct tenant. The software attaches tokens and attributed usage cost to each customer using auth-derived tenant context and stable event identity.
What are the core metering and control features?
The core metering features include ingesting events via SDK or HTTP Custody, pricing with model-specific input/output token rates, and attributing usage by customer, period, model, or span. Optional hard gates allow checking, reserving, and reconciling usage before it exceeds a prepaid or budget limit. These features are designed to prevent overspend on the hot path.
What does FluxMeter Intelligence do?
FluxMeter Intelligence provides heuristic and rule-based APIs under the /intelligence/* endpoint for root-cause decomposition and unit economics. It offers scenario simulation, pricing recommendations, profitability views, and forecast alerts. This intelligence layer reads native usage and cost rollups and can optionally overlay revenue data from OpenMeter.
How does FluxMeter compare to OpenMeter, Lago, or Langfuse?
The software complements other tools by focusing on runtime metering correctness rather than just invoicing. OpenMeter and Lago are cited for invoicing, while Langfuse and Helicone are used for observability. FluxMeter fills the gap by providing hard gates and customer-level attribution that these tools may not emphasize.
How do I integrate with the fewest lines of code?
Users can integrate with the fewest lines of code by using the Python SDK version 2.0.1. The wrap() method runs pre-call checks and post-call tracking, allowing configuration of fail_open behavior. Metering failures can invoke on_metering_error callbacks to handle issues gracefully.
How are token costs calculated and priced?
The pricing model uses model-specific catalog rates from engine config/pricing.json, such as input at $0.15/MTok and output at $0.60/MTok for gpt-4o-mini. The system calculates costs in real-time and returns a customer usage line with a cost_usd field, for example $0.00042 for 1,600 tokens.
Can I export usage data to a billing system or data warehouse?
The software includes optional export capabilities to push usage lines directly to a billing system or a data warehouse. This allows seamless integration with existing financial infrastructure without requiring a complete billing suite within FluxMeter itself. The export feature is part of the optional hard gates functionality.
How does it handle multi-tenant cost attribution?
The software can be used to attribute LLM token costs to each customer in a multi-tenant SaaS environment. It handles the complexity of provider invoices being account-level by tagging every token event to the right tenant. This ensures accurate cost allocation for billing and reporting purposes.
What ingestion methods are supported?
FluxMeter is designed to work with various upstream AI accounts and can be configured to handle different model-specific pricing. The system supports both SDK and HTTP Custody ingestion methods to accommodate different integration preferences. It is suitable for environments requiring precise token-level tracking.
How does the budget control mechanism work?
The software includes a check, reserve, and reconcile mechanism to control usage before it exceeds limits. This allows for proactive budget management and prevents unexpected overspend on AI API calls. The optional hard gates are a key feature for maintaining cost predictability.
What does a customer usage line look like?
The system provides a customer usage line that includes fields like bucket, input_tokens, output_tokens, total_tokens, and cost_usd. This structured output allows for easy parsing and integration with downstream billing or analytics systems. The example response demonstrates the format of the usage data.
How can I contact FluxMeter?
You can reach FluxMeter by e-mail at hello@fluxmeter.dev.
What is the official website of FluxMeter?
The official website of FluxMeter is fluxmeter.dev.
AI-generated from reviews on this page. Verify anything important with the company directly.
When comparing FluxMeter to comprehensive billing platforms, the distinction between core metering and full billing becomes apparent. FluxMeter focuses on ingesting usage events and attributing costs based on model-specific rates, which is efficient for tracking. However, the export functionality is listed as optional, meaning you may need to build your own pipeline to send usage lines to a warehouse. The supported stack uses Kafka and Flink for aggregation, which is robust but requires infrastructure knowledge. ClickHouse serves only as an audit cold store, not the billing truth. While the HTTP Custody and SDK options are helpful, the lack of a built-in billing suite means additional development work is likely required to achieve a complete customer billing solution compared to all-in-one alternatives.
I have been evaluating usage metering solutions for our multi-tenant AI gateway, and FluxMeter stands out as a practical choice. The availability of an open-source Python SDK and an OpenAPI specification makes integration straightforward. Having the ability to run the supported stack locally with a simple make command is incredibly convenient for testing before deployment. The architecture clearly supports customer-level token tracking without the overhead of a full billing suite, which is exactly what we needed. It seems reliable and well-documented for developers looking for transparency in their AI cost attribution.
The core metering logic looks solid for tracking tokens, but the pricing transparency leaves something to be desired. The example cost relies on catalog rates from an engine config file, which means you have to dig deeper to understand the actual numbers. For someone considering this company, it is unclear if the rates are fixed or if they change based on the model.
News
No news yet. This is where the business shares its own updates and press.
Photos
No photos yet. Photos shared by reviewers will show up here.
Widgets
Get the widget for your site →About FluxMeter
FluxMeter is an open-source usage metering tool designed for AI gateways and multi-tenant products. It tracks usage at the customer and model level, attributing token consumption and cost without providing a full billing suite. The software ingests usage events via an SDK or HTTP Custody endpoint, applies model-specific token pricing, and attributes usage by customer, period, model, or span. Optional features include budget control (check, reserve, reconcile) and exporting customer-level usage lines to a billing system or data warehouse. It is distributed under an Apache 2.0 license and includes a Python SDK, OpenAPI specification, and a runnable demo via a public GitHub repository.
- Website
- fluxmeter.dev
- hello@fluxmeter.dev