AI Cost and Token Consumption

Measure AI consumption across pull requests, developers, teams, models, and vendors using token usage data collected from supported AI providers.

LinearB presents AI consumption as an estimated cost in USD. This value is calculated from the models and tokens consumed during AI-assisted development.

Consumption (Estimated Cost) represents the value of AI resources consumed. It is not the amount billed by your AI provider.

The initial release supports:

Claude API integrations and OpenTelemetry
Cursor

GitHub Copilot support is planned for a future release.

Summary

Track AI consumption in LinearB as an estimated cost.
Analyze AI usage by users, teams, repositories, pull requests, models, and vendors.
Use predefined Custom Metric Builder reports for daily AI consumption and AI consumption per pull request.
Build custom reports using AI consumption and token usage data.

Overview

LinearB collects token usage from supported AI providers and correlates that data with engineering activity.

Consumption is shown as an estimated dollar value calculated from the AI models and tokens consumed. This provides a consistent way to compare AI usage across users, teams, repositories, pull requests, and providers, regardless of each provider's billing model.

Use Consumption (Estimated Cost) to understand how AI resources are used across your engineering organization and how that usage relates to delivered work.

Supported providers

Claude API — Consumption, token usage, AI models, skills, and tool calls.
Claude OpenTelemetry (OTEL) — Consumption, token usage, AI models, skills, and tool calls.
Cursor — Consumption, token usage, and AI models.

Available reports

Two predefined reports are available in the Custom Metric Builder.

Daily AI Consumption (Estimated Cost)

Daily AI Consumption (Estimated Cost) displays AI consumption over time using an estimated dollar value.

Use this report to identify changes in AI usage, monitor adoption trends, and compare consumption across time periods.

AI Consumption per Pull Request (Estimated Cost)

AI Consumption per Pull Request (Estimated Cost) displays the AI resources associated with individual pull requests.

Each row can include:

Pull request ID
Pull request title
Pull request URL
Total tokens
Consumption (Estimated Cost)

Use this report to identify AI-intensive implementation work and understand AI consumption by delivered feature.

Build custom AI consumption reports

AI consumption and token usage data is available in the Custom Metric Builder.

You can create custom reports such as:

Consumption by engineer
Consumption by team
Consumption by repository
Consumption by release
Consumption by AI model
Consumption by AI vendor
Daily AI adoption
Consumption by project

Leadership use cases

AI consumption data helps engineering leaders answer questions such as:

Which teams consume the most AI resources?
Which AI models are used most frequently?
Which pull requests required the most AI assistance?
How much AI consumption supports each delivered feature?
How does AI usage vary across teams, repositories, and projects?
Which AI vendors are used most across the organization?

Data collection

LinearB aggregates AI usage into daily per-user records and correlates that data with engineering activity.

AI usage can be correlated with:

Pull requests
Repositories
Commits
AI interactions
Developers
Teams

This allows AI consumption to be analyzed alongside existing engineering metrics and custom reports.

Limitations

GitHub Copilot support is planned for a future release.
Historical AI data is currently backfilled for up to 90 days.
Available metrics depend on the capabilities of each AI provider.
Skills, tool-call, and model analytics may not be available for all providers.
Consumption is an estimated value based on token and model usage. It does not represent provider billing or invoice totals.