See exactly what every request costs, which models burn the most budget, and where to cut spend. All on your own infrastructure.
LLM API bills are opaque. Provider dashboards show aggregate spend but not per-request breakdowns. You cannot tell which feature, which user, or which prompt pattern is driving costs. By the time you notice the bill, the damage is done.
Most cost tracking tools are SaaS products that require sending your prompts and completions to a third party. That creates a new data residency problem while solving the billing one.
Every request through the proxy is logged with input tokens, output tokens, model, provider, latency, and calculated cost in USD. The data stays in embedded SQLite on your server.
The Lookout dashboard shows cost breakdowns by model, provider, day, and user. You can set spend caps per team or per API key. The proxy tracks costs even in proxy-only mode with zero configuration.
Per-request cost with token-level granularity. Daily and monthly spend rollups per project. Cost breakdowns by model, provider, and user. Cache hit rates and cost savings from semantic caching. Historical cost trends with 7/30/90-day views.
All of this runs locally. No data leaves your network. No SaaS dependency. No per-seat pricing for your observability tool.
Stockyard does not just track costs. The Drover autopilot can automatically route requests to cheaper models when quality thresholds are met. Semantic caching deduplicates identical prompts. Spend caps prevent runaway costs before they hit your provider bill.
Aggregate cost dashboards tell you how much you spent last month. Per-request cost tracking tells you why. When you can see that a single customer's workflow generated $47 in API costs because of a retry loop, or that your embeddings pipeline is re-embedding documents it already processed, you can fix the problem instead of just observing the bill. Stockyard logs every request with its token count, model, provider, latency, and calculated cost in USD. The data lives in SQLite on your server, queryable through the API and visualized in the Lookout dashboard.
Cost attribution by API key lets you allocate spend to teams, projects, or customers. Set up separate keys for your production app, your staging environment, and your internal tools. The spend endpoint breaks down costs by key, by model, and by day. Export to CSV for finance reporting. Set spend caps that return a structured error when a key exceeds its budget — this prevents the runaway loop scenario where a single bug generates a four-figure bill overnight.