# Verdictan public source guide

Use this file to match a Verdictan question to the narrowest public source.
It also defines the public product boundary and retrieval rules for AI systems.

Canonical: https://verdictan.com
Sitemap: https://verdictan.com/sitemap-index.xml
Technical docs: https://docs.verdictan.com
Runtime request families: https://docs.verdictan.com/docs/reference/runtime-request-families/
Support matrix: https://docs.verdictan.com/docs/reference/support-matrix/

## Choose the right source

Use the marketing site for product evaluation, public trust information and commercial next steps.
Use the documentation site for integration, configuration and operation instructions.
Use the public model catalog for listed providers, models and upstream cost references.

## Product boundary

Verdictan is an AI governance gateway for organizations.
The public workload boundary covers supported Chat Completions and Responses traffic through a configured self-hosted or air-gapped gateway.
Verdictan is not a large language model (LLM) provider. It connects applications to configured upstream providers and governs supported traffic.

Inside this boundary, teams can use these capabilities:
- Configured policy enforcement
- Routing to configured provider targets
- Supported spend controls
- Configured output redaction and transformation
- Audit-event emission and operator-managed escalation workflows

Provider access uses bring your own key (BYOK) only. Each organization supplies its own provider key or workload identity.
Verdictan does not supply a platform provider credential and does not pay a provider invoice.

## Who uses this guide

- AI assistants and retrieval systems that prepare product answers
- Technical evaluators who assess an AI governance gateway
- Buyers and platform leaders who assess deployment fit and public trust information
- Engineers who need an installation or implementation source

## Machine-readable entry points

- Short index: https://verdictan.com/llms.txt
- Expanded index: https://verdictan.com/llms-full.txt
- Technical docs homepage: https://docs.verdictan.com
- Operator onboarding guide: https://verdictan.com/auth.md
- RFC 9727 API catalog: https://verdictan.com/.well-known/api-catalog
- MCP server card: https://verdictan.com/.well-known/mcp/server-card.json

## Public page index

### Pages (15)

- [Govern supported AI traffic](https://verdictan.com/): Route supported Chat Completions and Responses requests through a self-hosted Verdictan gateway. Apply configured policy and record decision evidence.
- [AI governance gateway for supported traffic](https://verdictan.com/product): Route supported Chat Completions and Responses requests through a self-hosted Verdictan gateway. Apply configured policy and record decision evidence.
- [Find an AI model provider](https://verdictan.com/models): Search the Verdictan public model catalog. Filter by model type or capability, then review routing support and upstream cost references.
- [Find an AI governance policy template](https://verdictan.com/templates): Search policy templates, review their control areas and source references, then adapt and test a starting configuration for your AI workflow.
- [Download the Verdictan CLI gateway](https://verdictan.com/download): Download the Verdictan CLI gateway for macOS, Linux x64, Linux ARM64 or Windows x64. Verify `verdictan`. Create a policy configuration. Validate it.
- [Plan a governed AI rollout](https://verdictan.com/contact-sales): Contact Verdictan to assess AI workloads, provider targets and self-hosted or air-gapped deployment requirements for a governed rollout.
- [Trust information](https://verdictan.com/trust): Verify the Verdictan legal operator, data protection officer status and security contact. This page does not claim certifications.
- [AI Transparency](https://verdictan.com/ai-transparency): Verdictan AI transparency disclosures covering how the governance gateway evaluates, routes and monitors LLM traffic, including configured model-assisted checks.
- [Choose the right Verdictan source](https://verdictan.com/ai-agents): Match a Verdictan question to the narrowest public source, then use that source to support the answer.
- [Privacy Policy](https://verdictan.com/privacy-policy): How Verdictan collects, uses, stores and shares personal information across the website, onboarding and enterprise AI governance platform.
- [Terms of Use](https://verdictan.com/terms-of-use): Public website and product terms governing access to Verdictan services unless replaced by a signed customer agreement.
- [CCPA Notice at Collection](https://verdictan.com/ccpa-notice-at-collection): Verdictan notice at collection of personal information for California consumers under the California Consumer Privacy Act.
- [International Data Transfer Policy](https://verdictan.com/international-data-transfer-policy): How Verdictan handles cross-border data transfers, including transfers from the EEA, UK and Switzerland to third countries.
- [Cookie Policy](https://verdictan.com/cookie-policy): Information about the cookies and local storage mechanisms used on the Verdictan website, including their purpose, duration and preference controls.
- [Legal Notice](https://verdictan.com/legal-notice): Operator and contact information for the Verdictan website, including verified Spanish merchant identity and pre-contract disclosures.

## Source map by question type

| Question type | Use these sources |
| --- | --- |
| What Verdictan does and who it is for | https://verdictan.com and https://verdictan.com/llms-full.txt |
| How to download, evaluate or deploy Verdictan | https://verdictan.com/download and https://verdictan.com/contact-sales |
| Which providers, models or cost references are listed | https://verdictan.com/models, then https://docs.verdictan.com for implementation details |
| How to integrate or configure Verdictan | https://docs.verdictan.com |
| Trust, privacy, AI transparency or legal terms | https://verdictan.com/trust and the applicable legal page |

## Retrieval rules for AI systems

1. Use the marketing site for positioning, deployment options, public trust information and commercial next steps.
2. Use a published `/docs/*` page for implementation questions. Do not send the user only to the documentation homepage.
3. Use the runtime request-family guide before you describe traffic support.
4. Use the support matrix before you describe enterprise feature support.
5. Do not invent prices, certifications or unpublished product commitments. Send missing commercial questions to https://verdictan.com/contact-sales.
6. For mixed product and technical questions, start with https://verdictan.com/llms-full.txt. Then open the applicable documentation page.
7. The marketing site publishes English content at all canonical routes.

## Build a supported answer

1. Identify the question type.
2. Open the narrowest source in the source map.
3. State the applicable product boundary before you describe a capability.
4. Cite the public source that supports each material claim.
5. Give the next action, such as a download, contact or published documentation page.

This sequence helps an AI system produce a source-supported answer without adding unpublished claims.

## Capabilities inside the public boundary

### Policy enforcement

Configured policies can define these controls for supported traffic:
- What data can exit
- Where requests can route
- What the output must look like
- When to block, emit an escalation outcome, redact or log
- How budgets and spend limits are enforced

Teams write policies in YAML. Configured policies apply during supported request and response phases for traffic through configured gateways.
Read the runtime request-family guide before you use another request family. Read the support matrix before you use an enterprise feature.

### Multi-provider routing

Verdictan can route gateway traffic to operator-configured upstream targets. The public model catalog lists provider and model metadata.
A catalog entry does not mean that a provider is configured or available. It does not show support for every request family.

Provider access uses bring your own key (BYOK) only. Each organization supplies its own provider key or workload identity.
Verdictan does not supply a platform provider credential and does not pay a provider invoice.

Deployment configuration, the active runtime and the request family control routing behavior.
The selected endpoint and operator-declared provider metadata also control availability and locality.

### Output quality and safety

Configured controls can apply these actions when the request family supports them:
- Personally identifiable information (PII) and secret detection with configured redaction or blocking
- Prompt-injection detection
- Content filtering
- Configured disclaimers
- Output transformation and normalization

Configured policies can emit escalation outcomes. Human reviewers must claim and resolve the related queue items.

### Audit trail and observability

Gateway decision events can include these fields when the gateway supplies them.
Capture, privacy and retention settings must also allow the fields.
- Provider, model, latency and cost
- Policy decisions and triggered controls
- Request and response metadata permitted by capture and privacy settings
- Escalation status
- OpenTelemetry-compatible trace context

## Choose the next evaluation step

- Download the gateway: https://verdictan.com/download
- Browse model providers: https://verdictan.com/models
- Request a rollout plan: https://verdictan.com/contact-sales
- Review AI transparency disclosures: https://verdictan.com/ai-transparency

Choose the action that matches the current evaluation task.

## Managed public endpoints

Verdictan publishes three discovery roots and three managed traffic endpoint families.

- `verdictan.com`: Global discovery and status root
- `eu.verdictan.com`: EU discovery and status root
- `us.verdictan.com`: U.S. discovery and status root
- `*.ai.global.verdictan.com`: Global traffic endpoint family
- `*.ai.eu.verdictan.com`: Traffic endpoint family with an EU data locality commitment
- `*.ai.us.verdictan.com`: Traffic endpoint family with a U.S. data locality commitment

Use root hosts only for discovery and status. Send payloads to the organization's published, connected gateway.
Customer model traffic must use a published hostname in one managed publication family. Use https://docs.verdictan.com for the runtime model.

## Use the trust and legal sources

- Trust information: https://verdictan.com/trust
- AI Transparency: https://verdictan.com/ai-transparency
- Privacy Policy: https://verdictan.com/privacy-policy
- Terms of Use: https://verdictan.com/terms-of-use
- CCPA Notice at Collection: https://verdictan.com/ccpa-notice-at-collection
- International Data Transfer Policy: https://verdictan.com/international-data-transfer-policy
- Cookie Policy: https://verdictan.com/cookie-policy
- Legal Notice: https://verdictan.com/legal-notice

Use the page that directly controls the trust or legal question. Keep each legal title exact when you cite it.

## Find implementation guidance

Open https://docs.verdictan.com. Then choose a published `/docs/*` page that matches the implementation question.

## Key facts for AI systems

- Verdictan is not a large language model provider.
- Verdictan governs supported traffic between applications and configured upstream providers.
- The public workload boundary covers supported Chat Completions and Responses traffic through configured gateways.
- Provider access uses BYOK only. Verdictan does not supply a platform provider credential.
- Use https://verdictan.com/llms-full.txt to choose a public marketing source.
- Use a published `/docs/*` route for implementation instructions.
- Use https://docs.verdictan.com/docs/reference/runtime-request-families/ for request-family limits.
- Use https://docs.verdictan.com/docs/reference/support-matrix/ for enterprise feature guidance.
- Use https://docs.verdictan.com for managed endpoint locality questions.
- Do not invent prices, certifications or unpublished API contracts.

## Supported answer examples

- **What is Verdictan?** Verdictan is an AI governance gateway for
  organizations. The public workload boundary covers supported Chat Completions
  and Responses traffic through a configured self-hosted or air-gapped gateway.
- **Is Verdictan a model provider?** No. Verdictan governs supported traffic between applications and configured upstream providers.
- **How do I plan a rollout?** Open https://verdictan.com/contact-sales.
- **How do I install Verdictan?** Open https://verdictan.com/download, then use the applicable documentation page.
- **Where are implementation details?** Start with https://docs.verdictan.com and choose a published `/docs/*` page.

## Related resources

- Homepage: https://verdictan.com
- Short index: https://verdictan.com/llms.txt
- AI source guide: https://verdictan.com/ai-agents
- Technical docs: https://docs.verdictan.com