oConsent
Consent infrastructure · SDK + API

Consent infrastructure for AI products.

Issue, verify, revoke, enforce, and audit machine-readable consent across datasets, agents, applications, and AI pipelines.

Built for AI-era data governance. Designed for LLMConsent-style consent records.

Start with SDK → Read docs
Explore LLMConsent ↗
  • TypeScript SDK
  • REST API
  • LLMConsent-compatible
consent_record rec_7f3a…e92
subject
user_123
asset
conversation_export
purpose
llm_training
actor
model_pipeline_7
issued
2026-06-28
expires
2027-06-28
proof
signed + timestamped
allowedsigned · revocable
verify.tspoint of use 
const consent = await oconsent.verify({
  subject: "user_123",
  asset: "conversation_export",
  purpose: "llm_training",
  actor: "model_pipeline_7"
});

if (!consent.allowed) {
  throw new Error("Consent not granted");
}
  1. Consent record Subject grants a scoped, signed permission.
  2. Verify Point-of-use check before any access.
  3. AI pipeline / agent Runs only on allowed assets.
  4. Audit proof Tamper-evident record of every use.
audit_trail 2026-06-28T10:22:04Z verify asset=conversation_export actor=model_pipeline_7 → ALLOWED
Fig. 1 / Point-of-use verification across an AI pipeline.
01 / The gap

Consent was built for banners. AI needs infrastructure.

01

Consent gets separated from data

A permission is captured in one system while the data flows into many. By the time a dataset reaches a model, the consent behind it is gone.

02

AI use is hard to prove

When a customer or regulator asks what a model was trained on, and why it was allowed, screenshots and spreadsheets are not evidence.

03

Revocation rarely propagates

Someone withdraws consent, but the data already sits in pipelines, caches, and fine-tunes that never hear about it.

02 / The model

Consent that travels with data.

One record, checked everywhere it matters. The same consent object follows your data from the moment it is issued through to the audit trail.

  1. 01

    Issue

    Mint a signed, machine-readable consent record that names the subject, asset, purpose, and allowed actors.

  2. 02

    Verify

    Check permission at the point of use with one call, before any dataset, agent, or pipeline touches the data.

  3. 03

    Enforce

    Wire enforcement hooks into your code so disallowed use is blocked by default, not flagged after the fact.

  4. 04

    Revoke

    Withdraw a permission once and let revocation propagate to every consumer that checks.

  5. 05

    Audit

    Produce tamper-evident audit proof of who used what, for which purpose, and under whose consent.

03 / Quickstart

Add consent verification in minutes.

Install the SDK, point it at your registry, and gate any AI workload on a live consent check. No banner, no bespoke plumbing.

npm install @oconsent/sdk
  1. Install the SDK
  2. Configure your API key
  3. Gate the workload on verify()
pipeline.ts@oconsent/sdk 
import { OConsent } from "@oconsent/sdk";

const oconsent = new OConsent({
  apiKey: process.env.OCONSENT_API_KEY
});

const result = await oconsent.verify({
  subject: "user_123",
  asset: "support_chat_2026_04",
  purpose: "model_finetuning",
  actor: "customer_support_llm"
});

if (result.allowed) {
  await runPipeline();
} else {
  await skipAsset(result.reason);
}
04 / The standard

Built for open AI consent standards.

LLMConsent defines what AI consent should mean. OConsent makes it enforceable in software.

The open standard

LLMConsent.org

  • Open standard
  • Public vocabulary
  • Consent specification
  • Governance
  • Ecosystem alignment
The infrastructure

OConsent.io

  • SDKs
  • APIs
  • Verification service
  • Consent registry
  • Enforcement hooks
  • Audit infrastructure
Explore LLMConsent ↗
Writing

Notes from the build.

Revocation and audit: making AI data use provable

Withdrawing consent so it actually propagates, and producing evidence of every decision that you cannot quietly edit later.

Point-of-use verification: gating AI workloads on consent

How to turn a consent record into an allow or deny decision at the moment data is about to be used.

The consent record: a machine-readable schema for AI data use

A field-by-field look at the consent record OConsent issues, and why each part is in there.

Read the blog →

Build consent-aware AI.

Start with verifiable consent records and make AI data use auditable by design.

Start with SDK →Read docs