Machine-readable product context

SpendLens AI product facts

SpendLens AI shows where your OpenAI and Anthropic money goes and where you can save without breaking your app.

Account Connect

Automatic OpenAI reporting through a dedicated Admin key.
Anthropic Cost and Usage CSV imports without an Anthropic API key.
Spend, model, token, project, workspace, API-key, caching, and batch insights when provider data supports them.
Separate financial views for each connected AI account.
Ownership mapping for applications, teams, and environments.

Runtime telemetry

Add SDK telemetry only when you need application-level attribution, prompt waste signals, and stronger evidence that a cost-efficient model can preserve quality.

Compare Mode

SpendLens shortlists up to three cost-efficient, policy-approved candidates for each classified workload.
Candidate calls run inside the customer's staging application with provider keys controlled by the customer.
The original model remains on the application response path. Candidate failures cannot replace its response.
SpendLens receives aggregate candidate cost, token, latency, reliability and error results.
Optional quality evaluation runs on SpendLens infrastructure using SpendLens provider credentials, a redacted template and safe generated data.
Dynamic customer values and production responses are not used for hosted quality evaluation.

Read the Compare Mode guide

Where SpendLens AI fits

Provider dashboards confirm native totals, observability tools help debug applications, and gateways route requests. SpendLens AI focuses on finding and validating AI cost optimization opportunities without becoming a proxy.

View alternatives and comparisons

Privacy

Account Connect does not read prompts or responses. Full user prompts are not sent by SDK telemetry by default. Template-only mode may use system prompts or templates; metadata-only mode disables prompt sampling. Candidate provider keys stay in the customer's staging application. Hosted quality evaluation does not use dynamic customer values or production responses.

Important qualification

Savings are estimates. Teams should evaluate quality, latency, safety, privacy, data residency, and provider agreements before migration.

Plain-text version: /llms.txt