You can install Muninark in your environment. The deployment plan defines source access, model routing and who operates each part of the system.
Your records, the Muninark installation and the model services you allow. The deployment plan names who runs each one.
Repositories, tickets, documents and other agreed sources.
Ingestion · knowledge stores · policies · MCP · operator and review tools.
Language models, embeddings, transcription and document or image parsing, where enabled.
Model, embedding, transcription and parsing services may run outside your environment. During evaluation, we identify which services receive data, where they run and which terms apply.
The engine provides these controls. Their effectiveness depends on the source metadata and deployment configuration, which need testing alongside your operating procedures.
The HTTP interface supports Entra OIDC token validation. Deployment policy maps caller entitlements to source scopes and sensitivity levels.
Configure the identity provider and access grants. With authentication disabled, the HTTP interface denies requests.
Retrieval and serving apply configured entitlement checks. Source references, coverage gaps and redaction indicators accompany supported responses.
Source metadata and policies determine the effective access boundary; they must be verified end to end.
The engine can store query outcomes in a hash-chained audit log and verify that chain. Several serving processes can share one log. Keyed chains and external checkpoints are optional.
Choose the sink, key management, retention and checkpoint storage for the deployment.
Edits at the source become new revisions. Content withdrawn or superseded at the source stops being served, and facts derived from it are flagged for expert review. Redaction and deletion mechanisms cover derived knowledge.
Deleting a whole source record and moving it between access scopes need an agreed procedure. Agree the scope across source data, derived records, indexes, logs and backups before making retention commitments.
A deployment plan covers the complete data path, including components operated by your team and external providers.
Hosting, network access, encryption and keys, backups, restoration and monitoring. Sizing follows corpus and workload measurements.
Provider accounts, model selection, data handling and usage budgets. Model changes need evaluation against the agreed questions.
Connector maintenance, source freshness, expert review, upgrades and incident escalation. Support terms are agreed in the engagement.
Start with the architecture and actual data flows. Agree the evidence your security team needs before access is granted.
This page describes the product’s control model. It is not a certification or a statement that a customer’s deployment satisfies a particular regulatory obligation.
Share your hosting and access requirements so we can work through the data flows and proposed setup.