Engineers first. Salespeople second.
Immutable audit trail
Insert-only logs with e-signatures and timestamps for legal and compliance review.
Agentic trust scoring
Live trust scores for AI agents detect scope drift, privilege escalation, and exfiltration.
Client-side PII interception
Intercept LLM prompts before egress to prevent leaks invisible to network DLP.
Everything you need to get Shield into production lives here — quickstarts, API references, deployment guides, and the policy cookbook our own engineers use. The fastest path is usually the gateway: point your AI traffic at Shield, apply a starter policy, and watch the first verdicts land in your SIEM within minutes.
If you would rather embed detection directly, the SDKs wrap the same engine behind a few lines of code in your language of choice, and the webhook reference covers wiring alerts into the tools your team already lives in. Prefer to read before you build? Start with the architecture overview, then the policy guide — together they explain not just how Shield works, but why each control exists.
Quickstart
Stand up a tenant, deploy the agent, see your first detection in 20 minutes.
Splunk recipe
OCSF events to HEC + 4 prebuilt dashboards: PII Firewall, Agent Trust, Deepfake, Coverage.
Custom patterns
Regex, LLM-classifier, semantic — and how to test them safely against synthetic prompts.
BYOK + HSM
Customer-managed keys via AWS KMS, GCP KMS, Azure Key Vault, or on-prem Thales.
Webhooks + API
REST + gRPC. Idempotency keys. Pagination. Rate limits documented.
SCIM + SAML
Provisioning + JIT. Group-mapped policy. Policy-as-code via Terraform.
# .defog/policy.toml
[siem.splunk]
hec_url = "https://splunk.corp/services/collector"
token_secret = "defog-prod/splunk-hec"
format = "ocsf"
[firewall.pii]
patterns = ["ssn_us", "pci", "phi_default", "corp_secret_internal"]
on_match = "block_and_redact"
escalate_to_classifier_on = ["corp_secret_internal"]
Have any questions?
Talk to the team that built Defog — security and product engineers, not a sales desk.