How NOFire AI protects your production and your data
Investigations only read. Every action passes your policy, and every action is recorded.
Every production action passes your policy and is recorded
Policy can require a person to approve any production action. The approver decides in Slack, and an agent cannot approve.
Set each agent and action to allowed, needs approval or denied, for each service and environment.
Every action is recorded: which agent asked, the context, what the policy decided, who approved, what ran and the result.
The audit log exports to your bucket every day. On Enterprise, agent activity also goes to your SIEM.
Certifications and security controls
Controls cover encryption, infrastructure, identity and process. The SOC 2 Type II report is available under NDA.
Audited annually against the SOC 2 Type II framework: security, availability, and confidentiality. Report available under NDA.
Request the report →TLS 1.3 in transit. AES-256 at rest. Customer-managed keys via KMS on Enterprise.
A short, published list of vendors. We notify you 30 days before adding a new one.
Hosted by NOFire AI, or in your cloud account on Enterprise
Both options run the same product with the same controls.
Multi-tenant and hosted by NOFire AI. The quickest way to start.
- Region-bound (US, EU)
- Tenant-isolated data planes
- Customer-managed keys (Enterprise)
- SOC 2 Type II · GDPR
NOFire AI runs in your cloud account, managed by NOFire AI.
- Deploys via Terraform / Helm
- Private connections only, no inbound
- Customer KMS & private registry
Sign in through your identity provider, with roles scoped to the work
Every person and every agent has an identity and a scope, and every action is recorded.
Running your own agents? Control Room runs each agent in a microVM, with policy checked outside the agent.
SAML SSO and SCIM provisioning on Enterprise, with Okta, Google, Azure AD, Auth0, OneLogin or Ping.
Map IdP groups to scoped roles.
Set each agent and action to allowed, needs approval or denied, for each service and environment. Read the runtime policy patterns →
Configurable session lifetimes, IP allowlists, and just-in-time elevation for sensitive scopes.
What NOFire AI stores, and where your data goes
Logs, metrics and traces stay in your tools. NOFire AI queries them when an investigation needs them and stores the findings, each query and a copy of each query result with credentials removed, for 90 days by default. NOFire AI also stores the production context built from your cloud, Kubernetes, code and CI/CD: services, environments, dependencies, deployments and changes. Your data is never used to train cross-customer systems.
Investigations use scoped read access to traces, metrics, logs and Kubernetes APIs. Actions use the access your policy grants, and every action is recorded.
Credential fields and token formats are removed from each query result before NOFire AI stores the copy. Stored copies are deleted after 90 days by default.
Your data is never used to train, fine-tune, or improve any cross-customer model.
Per-tenant Context & Control Model. No data, embeddings, or schemas are shared across customers.