Live demo below · Join the waitlist for API access

Immutable audit trails for AI decisions.
Prove your models are auditable in 30 minutes.

DecisionChain gives AI teams tamper-evident logs of every model decision — with bias detection and one-click evidence for the EU AI Act, ISO 42001, and SOC 2. Drop in the Python SDK, ship your models, hand auditors a link.

Early access: 50% off your first 6 months

View live demo →

Design partners get founder-level support. No spam — just launch updates.

decisionchain · chain verified ✓4 / 12,481
decision_idtimemodeloutputchain
  • dec_01HR9F4Q…10:14:22.481risk-scorer@3.2.1review · 0.82a91f…c204
  • dec_01HR9F4P…10:14:22.418risk-scorer@3.2.1approve · 0.11b03e…7d1a
  • dec_01HR9F4N…10:14:22.302risk-scorer@3.2.1approve · 0.24c8f2…9e40
  • dec_01HR9F4M…10:14:22.198risk-scorer@3.2.1review · 0.711d77…22ab
append-only · sha-256 chainanchored 10:15:00 UTC
audit.pypython
from decisionchain import DecisionChain

audit = DecisionChain(api_key=os.getenv("DC_KEY"))

decision = model.predict(input)

audit.log(
    model="risk-scorer@3.2.1",
    input=input,
    output=decision,
    actor=user.id,
)
01 · What it is

Compliance infrastructure that lives next to your model, not next to your policies.

DecisionChain is a drop-in SDK and hosted service that records every decision your models make into an append-only, cryptographically chained log. It runs bias and drift checks in the background, and turns those records into audit-ready evidence for the EU AI Act, ISO 42001, and SOC 2 — without your team building any of it in-house.

Immutable by design

Every record is hashed and chained to the previous one. Tampering breaks the chain and every downstream signature.

Bias & drift, watched

Continuous checks across your logged decisions, with thresholds and alerts you can point an auditor at.

Evidence on demand

Export a signed evidence pack scoped to a model, a customer, or a date range. Ready for regulators or buyers.

02 · How it works

Four steps from first log to audit-ready evidence.

  1. 01

    Install the SDK

    One pip install, one API key. Python-first, with a REST API for anything else.

    pip install decisionchain
  2. 02

    Log every decision

    Wrap your model call. Inputs, outputs, versions, and actors are captured with a single log() call.

    audit.log(model=..., input=..., output=...)
  3. 03

    Tamper-evident storage

    Records are hashed, chained, and anchored to an external witness. Alteration breaks the chain, provably.

    sha-256 chain · anchored hourly
  4. 04

    Export compliance evidence

    Generate signed evidence packs for the EU AI Act, ISO 42001, and SOC 2. Share a link, not a spreadsheet.

    audit.evidence.export(framework='eu_ai_act')

Evidence for the frameworks your buyers ask about

  • EU AI ActArticle 12 · logging
  • ISO/IEC 42001AI management system
  • SOC 2Type II · security
  • NIST AI RMFGovern · Measure · Manage

Framework references, not certifications. DecisionChain produces evidence your compliance team submits.

03 · Get access

Ship auditable AI before your first compliance conversation.

We're onboarding a small group of AI teams for early access. Design partners get founder-level support and 50% off the first 6 months. Drop your work email and we'll be in touch.

No spam. Just launch updates.