usertrust wraps any LLM client (Anthropic, OpenAI, Google) with a JS Proxy so every AI call becomes an immutable, auditable, double-entry financial transaction.

Keep your keys. Keep your billing. Add trust in one line.

import Anthropic from "@anthropic-ai/sdk";
import { trust } from "usertrust";

const client = await trust(new Anthropic(), { budget: 50_000 });

Core Guarantees

Two-phase spend lifecycle — Every LLM call follows PENDING hold → LLM call → POST (success) or VOID (failure). The same pattern banks use for credit card holds.
SHA-256 hash-chained audit trail — Append-only JSONL. Each event chains from the previous via deterministic canonicalization. Tamper-evident by construction.
Merkle proofs (RFC 6962) — Domain-separated hashing with inclusion and consistency proofs for public verifiability.
Real double-entry accounting — TigerBeetle ledger with 7 transfer codes. Not a counter.
Duck-typed client detection — Works with any Anthropic, OpenAI, or Google SDK. No direct imports.

Explore the Docs

Quickstart

Install and make your first governed LLM call in 2 minutes

Two-Phase Spend

How the banking hold pattern keeps your budget atomic

API Reference

trust(), defineConfig(), error classes, types

CLI Commands

init, inspect, health, verify, snapshot, tb

Policy Engine

12 operators, YAML rules, scope globs, time windows

Standalone Verifier

Zero-dependency vault verification

Stack

Layer	Technology
Runtime	Node 22, ESM
Language	TypeScript 5.9 (strict, noUncheckedIndexedAccess)
Ledger	TigerBeetle
Validation	Zod
Policy	YAML + minimatch
Test	Vitest (979 tests, 2.09:1 test-to-source ratio)

License

Apache 2.0 — View on GitHub

usertrust