Copy-paste recipes that run offline.

Instrument Chat Completions: budgets, recording, audit on the classic entrypoint.

The Responses API — the default for new OpenAI apps — through the same seam.

Budget + audit Claude calls, including prompt-cache read/write accounting.

Wrap the client under a LangChain/LangGraph agent; budgets and audit ride unchanged.

A Responses-based agent loop with pre-flight budgets and a tamper-evident trail.

contextkit packing alongside a RAG pipeline — with the assembly receipt.

Fully local, zero-key, end-to-end: a llama3 agent that's budgeted, recorded, audited.

A managed runtime owns the loop — ingest its gen_ai.* spans and keep budgets + audit.

A real agent test suite: record once, replay in CI, xdist-safe cassette layout.

Decisions, human oversight, policy flags → export an evidence pack and verify it offline.

The instrument() seam: one wrap, and every model call emits exact usage + decimal cost on the bus.

The same instrument() seam in TypeScript — exact token counts and decimal-exact money via @cendor/core.

Budgets, spend tracking, and reports — enforced pre-flight, before the call ever runs.

Assemble a token-budgeted context window with an auditable packing receipt.

Deterministic prompt compression with reversible, content-addressed handles.

Record a run once, replay it forever — deterministic, offline, free test fixtures.

A hash-chained audit log that verifies offline — one edited byte fails the chain.

The governed agent in TypeScript — zod tools, a pre-flight budget, and a tamper-evident audit chain on @cendor/sdk.