Skip to main content
Open source · Governix

Bring control, metering, and audit to your AWS Bedrock AI.

Governix is the open-source tenant-aware control, metering, and audit layer for AWS Bedrock AI and Bedrock RAG. Enforce runtime policy, track usage by tenant, and keep an audit-ready ledger of every request.

  • Enforce tenant-aware request policy before each Bedrock call
  • Track usage, quota, and estimated cost by tenant and application
  • Build an audit-grade ledger for retrieval and generation events

Governary is building the operating layer for teams turning Bedrock demos into real multi-tenant AI products.

HOW GOVERNIX FITS IN

Your Bedrock App

RAG application or AI feature

request

Governix — Policy Eval

Allow · Deny · Filter · Quota block

approved request

AWS Bedrock

Knowledge Bases · Models · RAG

response + events

Governix — Ledger & Usage

Audit · Quota · Showback · Export

Bedrock makes RAG easier to start — not easier to govern.

Once your AI feature serves multiple customers, the hard part is no longer retrieval quality.

01

Which tenant is allowed to access which knowledge base?

02

Which model should be used for this request?

03

What happens when a tenant exceeds quota?

04

How much did each tenant actually consume this month?

05

Can we explain how a specific answer was produced?

AWS Bedrock gives you the building blocks.

Governix gives you the governance layer.

Built for teams shipping Bedrock-based AI to real customers.

If you're moving from prototype to production, Governix is the layer you need.

B2B SaaS teams

You already have a Bedrock RAG feature or prototype, and now you need tenant-aware control, cost visibility, and auditability.

Platform and backend engineers

You need a clean way to enforce policy and collect runtime events without rebuilding your whole app around a gateway.

AI product and engineering leaders

You need to understand usage, quota pressure, and operational risk before your AI feature scales.

Security and compliance teams

You need request-level evidence, not just a general idea of what the system did.

Three things your Bedrock app needs in production.

01

Tenant-aware request policy

Define which knowledge bases and models a tenant can use. Apply runtime actions — allow, deny, force filter, downgrade model, or quota block — before each Bedrock call.

  • Tenant-specific policy rules
  • Runtime evaluation before Bedrock calls
  • Effective KB / model policy visibility
02

Per-tenant usage, quota, and showback

Track AI usage by tenant and application. Monitor request count, token usage, estimated cost, and quota thresholds.

  • Per-tenant request and token tracking
  • Quota thresholds and warning states
  • Showback views by tenant, model, and app
03

Audit-grade retrieval and generation ledger

Keep a request-level record so your team can investigate failures, explain decisions, and export audit evidence.

  • Request-level event history
  • Retrieval filter and chunk reference visibility
  • CSV export for audit workflows

Connect Bedrock RAG to Governix in five steps.

No proxy gateway required. Governix works as a control plane alongside your existing Bedrock integration.

01

Create a tenant

Represent each department, customer, or governed boundary as a tenant.

02

Register one or more applications

Each Bedrock RAG system connects as an application under a tenant.

03

Configure policy and quota

Define allowed knowledge bases, allowed models, citation requirements, fallback rules, and quota thresholds.

04

Integrate via SDK or middleware

Your existing Bedrock app calls runtime policy evaluation before the request, then sends runtime events after retrieval and generation complete.

Mode A · RetrieveAndGenerate Mode B · Retrieve + Converse
05

Operate from one control plane

Review usage, quota pressure, and audit ledger in the admin console.

No proxy gateway required for the MVP

Governix is designed as a control plane, not a mandatory inline proxy. Keep your existing Bedrock integration, add policy and audit incrementally, and avoid routing every model call through a new gateway on day one.

What the control plane shows.

A single console for tenant overview, usage tracking, and audit inspection.

app.governary.io/tenants

Tenants

4 active · 2 apps avg

Tenant Quota Status
AC

Acme Corp

tenant_01

72%
Active
GX

GlobalTech

tenant_02

89%
Warning
NV

NovaBio

tenant_03

31%
Active

Token usage · this month

per tenant

Audit ledger · recent

View all →
RetrieveAndGenerate
tenant_01
DENIED · quota_block
tenant_02
Retrieve + Converse
tenant_03
Open source first

Start with Governix, the open-source product from Governary.

The MVP validates three production needs — and nothing more. No billing platform, no multi-cloud support, no enterprise identity.

  • Can request policy be enforced before runtime?
  • Can tenant usage and quota be made visible?
  • Can every important request be explained and exported later?

MVP scope

Tenant management
Application registration
Runtime policy
Quota tracking
Audit ledger

Not in MVP

  • Billing platform
  • Multi-cloud support
  • Enterprise identity features

What Governix is — and where it fits.

Governix is
  • A tenant-aware AI control plane
  • A governance layer for Bedrock RAG
  • A usage, quota, and audit system
  • A runtime policy and event layer
Governix is not
  • A chatbot frontend
  • A vector database
  • A full billing platform
  • A general-purpose orchestration suite
Early access

Talk to the founder of Governary.

If you're building a multi-tenant Bedrock RAG product and already feel the pain of policy drift, quota uncertainty, or missing audit trails — let's talk.

Best fit for teams already running or piloting Bedrock RAG in a product setting.

Frequently asked questions.

What is the relationship between Governary, Governix, and Governair?
Governary is the website and umbrella brand. Governix is the open-source product. Governair is the planned cloud offering for the future.
Does Governix replace AWS Bedrock or Knowledge Bases?
No. Governix sits alongside your Bedrock-based RAG system as a control plane for policy, usage, quota, and audit. It does not replace Bedrock or Knowledge Bases.
Do I have to route every request through a proxy gateway?
Not in the MVP. The first version works through SDK or middleware integration plus runtime event ingestion. You keep your existing Bedrock integration and add governance incrementally.
Is policy defined per tenant or per application?
In the MVP, policy is tenant-scoped by default. Applications are used for runtime attribution, API key management, and usage breakdown.
What can I track per tenant?
Requests, token usage, estimated cost, blocked events, and request-level audit records — all broken down by tenant, application, and model.
Contact us

Tell us what you're building.

Share your Bedrock AI use case, current pain points, or what you want to evaluate with Governix.