token-accounting

Open measurement infrastructure for AI agents. Our audit of 110+ usage tools found 45+ verified billing bugs, with 23 fixes merged or accepted upstream (openlit, langfuse, codeburn, toktrack and more) and external real-world cases landing as conformance evidence. PASS / FAIL / UNPROVABLE in CI.

Bounded GenAI metrics and keyed summaries from OpenTelemetry traces

A prefix-cache profit-and-loss layer for DeepSeek coding agents — wrap your client in two lines, see per-request cache HIT/PARTIAL/MISS, the ¥ each prefix-bust wasted, and the reorder that restores the discount.

Deterministic benchmark for agent instruction adherence, with token and round-trip cost verified by a recording proxy

Local, deterministic, report-only Token ROI and multi-agent worktree accounting

Embeddable LLM benchmark speed measurement and offline Codex/Claude Desktop analysis, with explicit timing boundaries and output verification.

Conformance invariants and runnable harnesses for tools that count LLM tokens and cost from agent transcripts. Two catalogs, eleven upstream fixes.

Trace creator asset delivery token use and processing exceptions with one Java configuration.