Python Quickstart
Use the Pavri Python SDK to instrument an agent and send its registration, policy, and telemetry activity to the Pavri environment assigned to your organization.
Before you begin
- Python 3.12 or later
- A Pavri API key and organization identifier
- The gRPC endpoint and TLS requirement supplied by your Pavri administrator
- An agent object supported by one of the Pavri framework adapters, or a generic Python object
Do not substitute a guessed gateway URL for the gRPC endpoint. The SDK uses a
remote transport only when backend_grpc_url is configured; without it, the
SDK uses an in-memory client intended for local development and tests.
Install
python -m venv .venv
source .venv/bin/activate
python -m pip install --upgrade pip
python -m pip install pavri
Configure your environment
Store the values in your approved secret-management workflow, then expose them to the application process:
export PAVRI_API_KEY='provided-by-your-administrator'
export PAVRI_ORG_ID='your-organization-id'
export PAVRI_GRPC_ENDPOINT='provided-by-your-administrator'
Instrument an agent
import os
from pavri import PavriConfig, secure
class OrdersAgent:
name = "orders-agent"
def run(self, request: str) -> str:
return f"Processed: {request}"
secured_agent = secure(
OrdersAgent(),
PavriConfig(
agent_name="orders-agent",
plugin="generic",
api_key=os.environ["PAVRI_API_KEY"],
org_id=os.environ["PAVRI_ORG_ID"],
backend_grpc_url=os.environ["PAVRI_GRPC_ENDPOINT"],
tls_enabled=True,
environment="production",
team="platform",
),
)
print(secured_agent.run("refund request"))
For a local development endpoint without TLS, set tls_enabled=False only when
your deployment owner explicitly instructs you to do so. Production endpoints
should use TLS.
Verify the integration
- Run a representative agent request.
- Open Operate → Agents in Pavri and search for
orders-agent. - Open the agent and inspect its recent session or activity evidence.
- If it does not appear, first verify the organization ID, endpoint, network reachability, and credential scope with your administrator.
After confirming visibility, use policy authoring to introduce governance controls with a narrow, observe-oriented rollout.
Framework adapters
Pass the appropriate plugin value when securing a supported framework:
| Framework | Plugin value |
|---|---|
| LangGraph | langgraph |
| CrewAI | crewai |
| OpenAI Agents SDK | openai_agents |
| Claude SDK | claude |
| Google ADK | google_adk |
| AutoGen | autogen |
The wrapping pattern is the same: create the framework agent first, then pass it and a PavriConfig to secure().