Skip to main content
When AI agents take actions that cost or earn money, two questions become unavoidable: which decision caused which outcome, and can you prove it? Precipiq answers both. Every agent decision is written to an append-only, hash-chained ledger, then linked to the downstream financial event it caused — revenue earned, cost incurred, or liability opened. The result is a forensically auditable record from “the model said X” to “the customer was charged Y,” queryable years after the fact.

​
Why teams use it

Forensic traceability

Every decision is chain-linked and signable. Export a cryptographically sealed bundle for auditors, regulators, or post-incident review.

AI P&L

A live dashboard that attributes revenue, cost, and open liability to specific agents — so you can answer “is this AI making or losing money?”.

Compliance-ready

Designed to align with EU AI Act Article 12 record-keeping and the kinds of evidence U.S. class-action discovery tends to demand.

Drop-in SDKs

Python and TypeScript SDKs with LangChain, CrewAI, OpenAI, and Vercel AI adapters. Wrap a function; never touch transport code.

​
The 60-second mental model

Precipiq is built around four concepts that compose into a complete audit trail.
1

Decision

Your AI picks an action. The SDK writes a Decision Record: inputs, outputs, confidence, human-in-loop flag, and timestamp.
2

Chain

Each record hashes its predecessor’s hash into its own prev_hash field — tamper-evident and verifiable in constant time.
3

Financial event

A payment clears, a refund fires, an invoice books. Stripe and QuickBooks webhooks feed these in automatically; you can also POST events directly.
4

Consequence link

The attribution engine (or you) connects decisions to events. Live AI P&L updates. Threshold-exceeded liabilities fire dashboard alerts.

​
Where to go next

Quick start

Install the SDK, log your first decision, and see it on the dashboard in five minutes.

Core concepts

Understand Decision Records, Financial Events, Consequence Links, and the Hash Chain.

API reference

Full endpoint reference with interactive try-it forms and multi-language code samples.

Stripe integration

Auto-create Financial Events from Stripe webhooks and let the attribution engine propose consequence links.

​
Not a fit (yet)

Precipiq is built for product-side AI where an action costs money. It is not a model-observability replacement — tools like LangSmith and Arize are better suited for tracing prompts, inspecting activations, and profiling latency. Precipiq records decisions so the downstream consequences can be measured and defended, not to prevent bad decisions from happening.