FriskAI
Product

The policy engine for agent actions

FriskAI monitors tool calls made by LLM agents and evaluates them against deterministic policies before any action runs. It gives teams a structured way to understand and control how agents interact with real systems, without changing how they build their agents today.

Real time policy evaluation
Framework Agnostic
Structured observability

Core Capabilities

Everything you need to govern agents at scale. From policy modeling to deep observability.

Developer First
Zero-friction Developer Experience

Drop in the FriskAI SDK and get real-time visibility, enforcement, and auditability across every agent action.

  • TypeScript & Python SDKs
  • Framework adapters
  • Sub-ms latency
from friskai import FriskClient
from friskai.adapters.langgraph import FriskLangGraphHandler
from my_agent.graph import graph 

frisk = FriskClient(
    api_key='FRISK_API_KEY',
)

# install once
app = FriskLangGraphHandler(graph, frisk)
Simple & Powerful
Policy modeling

Express complex business rules simply.

  • Visual policy builder
  • JSON export
  • Version history
Policy Preview
{
  "enforce": [
    {
      "if": "amount > 10000",
      "then": [
        "block",
        { "notify": "slack://risk-team" }
      ]
    }
  ]
}
Full Visibility
Observability & Traceability

See exactly when your agents made a tool call along with the full trace of what happened before and after.

  • Complete traces of every agent action
  • Full context around each tool call
  • Step-by-step history before and after execution
Live enforcement feed
Secrets detected in tool input
no-secrets-prod
Blocked
api_key → ***********
Knowledge base lookup
customer-support-runtime
Allowed
Within tenant scope

Enterprise Readiness

Built for scale, security, and compliance from day one.

Audit logs
Permission models
Human-in-the-loop workflows
SOC2 readiness
SSO & SCIM provisioning
Incident escalations
Ready to ship governance

See the platform in action

Instrument a sandbox agent, review policy templates, and walk through a live incident with our team.