AI Vyuh ProvenanceOps
aivyuh provenance
FinOps watches what your AI costs. ProvenanceOps watches what your AI cites.

The AI provenance platform for production AI

Continuous monitoring of every source your AI cites. Detect stale knowledge, drift, and cross-source contradictions before they reach your users — and produce audit-grade evidence packs for AI Act, DPDP, FINRA, and HIPAA reviews. Metadata-only by default.

Confluence, Notion, SharePoint, RAG vectors
Metadata-only — never reads your prompts
KMS-signed evidence packs (Tier 3)
Free Provenance Audit — no signup

Knowledge Decay Index

0.38

+0.07 vs last month

Decay Flags

312

across 4 sources

High-Severity

7

resolved in <24h

Attribution Coverage

96.4%

board-ready metric

Daily Decay-Flag Volume

Flags High-severity
The Problem

Your AI sounds confident.
Its sources are out of date.

You wouldn't ship a database without monitoring writes. Why are you shipping an AI assistant without monitoring what it cites?

Stale Knowledge in Production

Your assistant is citing a deprecated policy from 2024. Customers got the wrong answer. Nobody noticed until the screenshot hit Twitter.

47% of AI answers reference stale docs

Cross-Source Contradictions

Confluence says one thing, Notion says another, and the RAG vector store has both indexed. Your AI picks whichever it retrieves first.

1 in 6 queries hit contradicting sources

Regulator Asks: Can You Prove It?

EU AI Act Article 10 requires data quality and currency evidence. DPDP, FINRA, HIPAA need audit trails for every cited fact. Screenshots aren't evidence.

€15M+ AI Act fine schedule per incident

Silent Knowledge Drift

Newer docs supersede older ones. Authors leave. Tags change. Your AI keeps quoting the old version because nobody told it to stop.

30 days median time-to-detect manually

What teams do today vs. what ProvenanceOps enables

Quarterly QA review

An analyst spot-checks 50 AI answers a week. Stale doc lands in the other 50,000.

Generic LLM observability

Datadog and friends track latency and cost. They don't tell you the source you cited was deprecated 90 days ago.

ProvenanceOps (V8)

Continuous source-attribution monitoring, decay detection, drift forecasting, and KMS-signed evidence packs. Provenance-native.

How It Works

5 minutes to connect.
Provenance every hour after that.

Connect your knowledge sources. Our agents handle the rest — monitoring, analyzing, and producing evidence continuously.

01

Connect Your Knowledge Sources

Point ProvenanceOps at your Confluence space, Notion workspace, SharePoint site, Glean audit log, or RAG vector store. Read-only metadata access — we never read document bodies.

P0: Confluence, Notion, Pinecone, Weaviate, pgvector, Chroma. P1: Glean, Notion AI, SharePoint via MS Graph.

5 min
02

Continuous Provenance Capture

Per-source schedulers fan out on tier-appropriate cadences. Confluence/Notion/SharePoint/Drive every hour. Glean every 6h. RAG vectors and web sources daily. All metadata normalized to a single SourceRecord schema.

Pathway 1: V8-pulled scheduled metadata. Pathway 2: customer-pushed telemetry via @aivyuh/provenance SDK.

Continuous
03

Intelligence & Analysis

Eight rule-based agents score provenance, flag decay, project drift, reconcile cross-source conflicts (metadata-only), assemble evidence packs, and reconstruct incidents on demand. LLM fallback hard-capped at <1% of records.

5 analysis categories: source provenance, decay detection, drift forecasting, cross-source reconciliation, audit & compliance.

Daily
04

Reports, Alerts, and Evidence

Dashboard, Slack/email/webhook alerts, monthly customer reports (PDF), quarterly executive board reports, and KMS-signed evidence packs that a third-party verifier can validate without our secret.

Tier 3 Regulated maps evidence to AI Act, DPDP, FINRA, and HIPAA clauses out of the box.

Weekly + on-demand

5 Intelligence Categories

C1

Source Provenance

Origin, lineage, freshness scoring

C2

Decay Detection

Stale, deprecated, superseded

C3

Drift Forecasting

30/60/90-day projections

C4

Reconciliation

Metadata-only conflict pairs

C5

Audit & Compliance

KMS-signed evidence packs

Developer SDK

Three lines of code.
Provenance on every response.

Install the SDK. Wrap your AI client. Every response is automatically tagged with the source(s) it cited — schedule-driven monitoring on top of that fans out across your knowledge bases.

TypeScript / Node.js @aivyuh/provenance
View on npm →
$ npm install @aivyuh/provenance
import Anthropic from "@anthropic-ai/sdk";
import { wrapClient } from "@aivyuh/provenance";

const client = wrapClient(new Anthropic(), {
  telemetryEndpoint: "https://provenance-api.aivyuh.com/provenance/telemetry",
  customerId: "your-customer-id",
  project: "support-bot",
  tags: { source_system: "confluence", document_class: "policy" },
});

// Use the client exactly as before — all types preserved
const message = await client.messages.create({
  model: "claude-sonnet-4-6-20260320",
  max_tokens: 1024,
  messages: [{ role: "user", content: "What is our refund policy?" }],
});

Two Pathways

Pathway 1: V8 pulls source metadata on schedule. Pathway 2: your client pushes per-response provenance via SDK.

Metadata Only

Captures source_system, source_doc_id, last_modified, author, citation tags. Never touches prompts or responses.

Fire-and-Forget

Telemetry runs in the background. If our endpoint is down, your app doesn't notice.

Verifier Package

@aivyuh/provenance-verifier validates KMS-signed evidence packs without any V8 secret. Audit-ready.

Python SDK is on the roadmap (post-Gate); design-partner Python stacks get early access. Email provenance@aivyuh.com.

Pricing

One avoided incident pays for the year

Start with a Free Provenance Audit. Upgrade when you see what's stale. Every paid tier has 93%+ gross margins because the agents are 95%+ rule-based, not LLM-billed per call.

Free

Provenance Audit
$0

One-time scan of one knowledge base. Branded PDF emailed in minutes — no subscription.

  • 1 audit per email per 30 days
  • Connect 1 source: Confluence, Notion, Glean, or up to 10K RAG docs
  • Knowledge Decay Index (KDI) score
  • Top-20 stale-flag candidates
  • Comparison vs public KDI baseline
Join Audit Waitlist

Starter

Tier 1
$50 /month

Continuous provenance monitoring for one production AI. Catch the stale-doc citation before your customer screenshots it.

  • 1 project, 1 source system
  • 30-day decay-flag history
  • Source provenance scoring
  • Decay alerts (dashboard)
  • Monthly customer report
  • Connectors: Confluence, Notion, RAG vectors
Get Started
Most Popular

Team

Tier 2
$300 /month

Full intelligence for AI Ops teams. One avoided wrong-answer incident pays for the entire annual subscription.

  • Up to 5 projects, 5 source systems
  • 90-day decay-flag history
  • Everything in Starter
  • Drift forecasting (30/60/90)
  • Cross-source reconciliation (metadata)
  • Slack / email / webhook alerts
  • All P0 + P1 connectors (incl. SharePoint, Glean)
  • Email support (<24h SLA)
Start Team Plan

Enterprise

Tier 3 Standard
$1,500 /month

Audit-grade evidence packs for non-regulated mid-enterprise. Board-ready provenance posture without the regulator-mapping premium.

  • Unlimited projects + source systems
  • 1-year decay-flag history
  • Everything in Team
  • KMS-signed evidence packs
  • Per-customer KMS asymmetric keys
  • Quarterly executive board report
  • Slack channel support (<4h SLA)
  • SSO / SAML
Contact Sales
Audit-Grade

Regulated

Tier 3 Regulated
$2,500 /month

Audit-grade + regulator mapping. The 1.67× lift covers AI Act / DPDP / FINRA / HIPAA clause mapping and dedicated regulator-comm support.

  • Everything in Tier 3 Standard
  • AI Act, DPDP, FINRA, HIPAA clause mapping
  • Customer-segment-aware severity rubrics
  • 1-min Decay Alerter cadence
  • Incident Reconstruction (on-demand)
  • Quarterly business review with Atin
  • Regulator-comm support
Contact Sales

20% off with annual billing. Early-adopter pilot: 50% off for 3 months (first 5 paying customers, any tier).

All prices in USD. INR pricing: ₹0 / ₹4,000 / ₹24,000 / ₹1,20,000 / ₹2,00,000 per month.

AiVyuh AI Operations Suite

FinOps and ProvenanceOps,
on one invoice and one login.

Run V7 FinOps and V8 ProvenanceOps side by side and your lower-priced subscription drops 25%. The higher-priced subscription stays at full list — so no tier path is ever cheaper inside the Suite than that tier is on its own.

How the Suite discount applies

1. Higher-priced subscription

Charged at full list. Always. The Suite never makes the bigger product cheaper than buying that bigger product alone.

2. Lower-priced subscription

Charged at 25% off. This is the only line the Suite discount touches.

3. No combined-sum math

We don't add the two prices together and then discount. That structure cannibalizes mismatched-tier pairs. We don't allow it.

V7 Tier 1 + V8 Tier 1

Engineering-led startup

V7 FinOps Tier 1

Higher-priced line · full list

$50/mo

V8 ProvenanceOps Tier 1

Lower-priced line · 25% off

$37.50/mo

$50/mo

Suite invoice

$87.50/mo

Two-product evaluation at the developer entry point. Higher-priced line stays at $50. Lower-priced line takes the 25% — V8 Tier 1 becomes $37.50.

V7 Tier 2 + V8 Tier 2

Mid-market AI Ops team

V7 FinOps Tier 2

Higher-priced line · full list

$300/mo

V8 ProvenanceOps Tier 2

Lower-priced line · 25% off

$225/mo

$300/mo

Suite invoice

$525/mo

Cost telemetry plus currency telemetry on a single dashboard. Higher-priced line stays at $300. Lower-priced line takes the 25% — V8 Tier 2 becomes $225.

V7 Tier 3 + V8 Tier 3 Regulated

Regulated enterprise

V8 Tier 3 Regulated

Higher-priced line · full list

$2,500/mo

V7 FinOps Tier 3

Lower-priced line · 25% off

$1,500/mo

$2,000/mo

Suite invoice

$4,000/mo

Article 10 of the AI Act + Article 13 transparency in one stack. Higher-priced line (V8 Regulated) stays at $2,500. Lower-priced line takes the 25% — V7 Tier 3 becomes $1,500.

V7 Tier 3 + V8 Tier 1

Mixed-tier pairing

V7 FinOps Tier 3

Higher-priced line · full list

$2,000/mo

V8 ProvenanceOps Tier 1

Lower-priced line · 25% off

$37.50/mo

$50/mo

Suite invoice

$2,037.50/mo

FinOps-led buyer adding V8 as a developer trial. Higher-priced line stays at $2,000 — never cheaper than V7 Tier 3 alone. Lower-priced line takes the 25% — V8 Tier 1 becomes $37.50.

Eligibility

Active paid subscription on both V7 and V8 — any tier combination, matched or mismatched. Free tiers do not unlock the Suite discount.

Stacking with annual

The Suite 25% on the lower-priced subscription applies first, then the 20% annual discount applies to whatever remains on each line.

FAQ

Questions buyers actually ask

What's the difference between FinOps and ProvenanceOps?

FinOps watches what your AI costs — token spend, model attribution, anomaly alerts on the bill. ProvenanceOps watches what your AI cites — source freshness, knowledge decay, cross-source contradictions, audit-grade evidence. They are two halves of AI Operations. The AiVyuh AI Operations Suite runs both; see the Suite section for how the discount applies to the lower-priced subscription.

Does ProvenanceOps read my prompts or AI responses?

No. The default posture is metadata-only. We capture source identifiers, last-modified timestamps, authors, tags, and citation links — not document bodies and not the natural-language responses. Tier 3 customers can opt into a separate semantic-comparison contract for specific reconciliation use cases; that is never the default and never silent.

What's a Knowledge Decay Index (KDI)?

A composite score from 0 (current) to 1 (heavily decayed) computed across all of a customer's monitored sources. It rolls up source freshness, deprecation markers, supersession signals, and metadata-only contradiction signals. We also publish a public KDI benchmark that ranks decay rates across major AI assistants — your Free Provenance Audit PDF compares your KDI against that public baseline.

Which knowledge sources do you support?

P0 (live): Confluence (REST v2), Notion, and four RAG vector stores — Pinecone, Weaviate, pgvector, and Chroma. P1 (live): Glean audit log, Notion AI, and SharePoint via MS Graph. Web sources are pulled daily. New connectors ship continuously; ask about your stack at provenance@aivyuh.com.

Why are there two Tier 3 SKUs?

Tier 3 Standard ($1,500/mo) is for non-regulated mid-enterprise that wants audit-grade evidence packs without the regulator-mapping premium. Tier 3 Regulated ($2,500/mo) adds AI Act / DPDP / FINRA / HIPAA clause mapping, customer-segment-aware severity rubrics, 1-minute Decay Alerter cadence, regulator-comm support, and a quarterly business review. The 1.67× lift is bounded by what regulated buyers actually pay for compliance tooling at this category — Drata and Vanta enterprise sit in the $2K–$5K range.

How does the Suite bundle pricing work?

The Suite charges your higher-priced subscription at full list and applies a 25% discount to your lower-priced subscription only. We do not sum the two prices and discount the total. That structure would let some mismatched-tier pairs end up cheaper inside the Suite than the bigger product is on its own — see the worked examples in the Suite section. The lower-priced-line approach is verified safe across every tier combination.

What's in the audit-grade evidence pack?

Every Tier 3 customer gets per-tenant KMS asymmetric signing keys. Evidence packs bundle decay-flag history, source-metadata snapshots, and reconciliation results, sign the bundle with your key, and write it to S3 with a DynamoDB index entry. A standalone @aivyuh/provenance-verifier package validates the signature without any AI Vyuh secret — your auditor can independently verify the pack.

What does the Free Provenance Audit actually do?

Connect one Confluence space, Notion workspace, Glean audit log, or up to 10K RAG documents. We run a one-time scan, score the corpus on the Knowledge Decay Index, surface the top-20 stale-flag candidates, and email a branded PDF that compares your KDI against the public baseline. No subscription. One audit per email per 30 days. Tokens encrypted at rest, deleted after the run. No prompt content stored.
Get Started

See what's stale in your AI's sources

Run a Free Provenance Audit on one knowledge base. Branded PDF emailed to you with a Knowledge Decay Index score and the top stale-doc candidates. No subscription. No credit card.