Runs in your infrastructure
Self-hosted by design. The document index, the embedding and language models, and the retrieval pipeline all run where you deploy KnowWiz. Nothing is sent to a vendor cloud.
Security and architecture
Your knowledge never leaves.
KnowWiz runs entirely in your own infrastructure. The document index, the models, and the retrieval pipeline stay inside your VPC, on-prem, or air-gapped network. Answers come only from documents the person asking can already open.
- Runs in your VPC, on-prem, or air-gapped
- Answers only from documents the asker can open
- Your models, not a third party you did not choose
KnowWizSecured
Deploy
runs in your VPC or air-gapped
Identity
OIDC / SAML SSO connected
Access
source permissions enforced
Audit
every query recorded
Security capabilities
Built to pass a security review.
Permission-aware answers
Retrieval enforces document-level access controls on every query. People get answers only from documents whose access includes them, so search never becomes a way around your controls.
Your models, your choice
Connect Anthropic, OpenAI, or Gemini, or run fully self-hosted models through Ollama or vLLM. Prompts and documents go only to the provider you pick, or to no external provider at all.
Traceable by design
Query history records who asked what and which sources an answer drew from. Role-based controls govern users, agents, connectors, and admin actions.
Sensitive data tokenized in the pipeline
Detected sensitive values are replaced with reversible, format-preserving tokens before a document is indexed and before a prompt reaches a model, so raw identifiers never sit in the index or leave for inference. Runs inside your boundary.
How it holds up
The architecture behind the claims.
Runs in your own environment
Deploy KnowWiz in your VPC, in your own data center, or in an air-gapped network with no outbound internet. The index, models, and pipeline run on infrastructure you control, so your documents and the queries against them never leave your boundary.
- VPC, on-prem, or fully air-gapped deployment
- Index, embeddings, and models run in your environment
- No document or query content sent to a vendor cloud
- You own the backups, the logs, and the data lifecycle
Accountability
Every question leaves a record.
AI search is only trustworthy if you can see how it was used. KnowWiz keeps a query history of who asked what, when, and which sources each answer drew from, so usage stays reviewable and answers stay traceable to their evidence.
- Query history for each user and agent
- Answers link back to the exact source passages
- Role-based controls over connectors and admin actions
- Logs stay in your environment for your own review
Honest comparison
Where a self-hosted layer differs.
How KnowWiz compares with a public AI chatbot and a typical hosted AI search product. KnowWiz is pre-launch, so this reflects how it is built, not third-party test results.
| KnowWiz (self-hosted) | Public AI chatbot | Typical hosted AI search | |
|---|---|---|---|
| Runs in your own infrastructure | Yes | No | No |
| Air-gapped deployment option | Yes | No | No |
| Answers respect source permissions | Yes | No | Partial |
| Choose or self-host your own models | Yes | No | Partial |
| Query history and audit trail | Yes | No | Partial |
| Single sign-on via OIDC or SAML | Yes | No | Yes |
| Every answer cites its sources | Yes | No | Partial |
Data protection, demonstrated
A typed identifier, tokenized in the trace.
A question containing the synthetic PAN ABCDE1234F was sent to the product. The reasoning trace shows the model working with the token IN_PAN_79c7ca06cce9 instead, detection logged, raw value never stored or sent.
app.knowwiz.ai
Security questions
What buyers ask before a pilot.
Everywhere KnowWiz runs is infrastructure you control. The document index, the models, and the retrieval pipeline sit inside your VPC, your data center, or an air-gapped network. Documents and queries do not leave that boundary.
Yes. KnowWiz supports a fully air-gapped deployment with self-hosted models, so it works with no outbound connectivity. Connectors then sync from internal systems only.
No. Retrieval enforces document-level access controls at query time. People only get answers from documents whose access includes them, and restricted content stays out of their results.
You choose the models: Anthropic, OpenAI, or Gemini through your own keys, or self-hosted models through Ollama or vLLM. Prompts and context go only to the provider you configure, and self-hosted models keep inference inside your network. KnowWiz does not use your content to train its own models.
Sign-in runs through your OIDC or SAML identity provider, so access follows your existing joiner and leaver process. Role-based access control governs who can build agents, manage connectors, and reach admin settings.
Query history records who asked what and which sources each answer drew from. The logs stay in your environment, so your own team can review usage and trace any answer back to its evidence.
Not yet. KnowWiz is pre-launch and has no certifications to claim. The architecture is designed to support your compliance review: self-hosting, permission-aware retrieval, SSO, role-based access control, and audit logging. We are glad to walk your security team through the design.
See it in your environment
Run the pilot inside your own boundary.
A working session on your infrastructure: connect a source, watch the index build, and ask real questions, all without a document leaving your network.
Self-hosted pilot on your infrastructure. Your data never touches ours.