Integration Guides

Nomotic integrates with any AI agent framework through the GovernedAgentBase class. All integrations follow the same pattern:

1. govern_action() — evaluate the action before execution

2. Execute — run the action if governance allows it

3. validate_output() — validate the agent's output after generation

The governed_run() method combines all three steps into a single call.

Available Integrations

Framework	Guide	Pattern
GovernedAgentBase	Core governance wrapper	Base class for all integrations
LangGraph	State graph with governed nodes	Governance as graph nodes with conditional edges
CrewAI	Governed crew tasks	Task wrapper with veto handling
OpenAI	Function calling governance	Tool dispatch with governance gate
Claude	Anthropic SDK integration	Governed API calls with output validation

Prerequisites

All integrations require:

pip install nomotic

Plus the framework-specific package (e.g., langgraph, crewai, openai, anthropic).

Core Pattern

Every integration follows this structure:

from nomotic import GovernanceRuntime
from nomotic.governed_agent import GovernedAgentBase

runtime = GovernanceRuntime()
# ... configure runtime with scope, rules, etc.

agent = GovernedAgentBase(runtime=runtime, certificate=cert)

# Option 1: Manual control
result = agent.govern_action("read", "customer/data")
if result.allowed:
    output = do_something()
    validation = agent.validate_output(output, action_type="read")

# Option 2: Full lifecycle
output = agent.governed_run(
    action_type="read",
    target="customer/data",
    execute_fn=lambda: do_something(),
)

Start with GovernedAgentBase to understand the core API, then see the framework-specific guides.