Architecture

The Operational Control Plane

Kindo builds, executes, and governs AI agents. The competitors scan and monitor them. Two different layers — not competitors.

Hard moat
No competitor offers this
On-prem / air-gapped
Runs inside the customer environment
Agent execution engine
General-purpose, any workflow
VMI sandbox
SQL, Vector DB, Object Store
Turbo Mode
Patterns compiled to deterministic code
Deloitte-custom build
CDA-native
Kindo
Operational control plane
inside → out
01 Inference proxy Every model call passes through, with DLP applied inline
02 MCP gateway Tool access brokered under RBAC per team and role
03 Telemetry & audit Full execution trail, exportable to the customer's SIEM
04 Governance & policy Client-configurable rules enforced at run time
Policy
DLP
Audit logs
Secrets manager
Capabilities
Integrations and outputs
200+ MCP servers
Connected tools and systems
26+ LLM models
Routed through one proxy
SIEM / SOAR export
OTel audit trail to the customer's stack
Governance dashboard
Usage, cost and policy in one view
Anomaly detection
From execution context, not network traffic
Complements, does not compete
Security inspection layer
Partner options · outside → in
Prisma AIRS · CrowdStrike · Google SecOps · Microsoft
Shadow AI discovery
Safety
Model scanning
Runtime security
Automated red teaming
Runtime security
Network intercept
Identity
Workload coverage
Observability
Categories per the Gartner Guardian Agent framework: safety, runtime security, identity, governance, observability.