Custom Python integration

Action security for Custom Python.

Call the Action Firewall before your own dispatcher executes.

Get the package and source.

Connect the execution boundary.

Your application submits the final tool arguments, rejects non-allow decisions and applies sanitised arguments when returned. Test errors and timeouts explicitly.

An example to test

A custom CRM agent attempts a record update beyond the requested task. Evaluate the final update and stop the dispatcher when policy denies it.

Illustrative workflow, not a customer deployment or a measured result.

Before connecting live policy

  1. Identify the handler that actually performs the side effect.
  2. Provide trusted identity, purpose and final action arguments.
  3. Configure the tenant’s assigned runtime and keep API credentials server-side.
  4. Test allowed, blocked and unavailable paths before production.

Integration example

Adapt this existing SDK pattern to your application. Names such as the agent, handler and session refer to your own objects. Follow the package documentation for compatible versions and complete setup.

# The core package has no framework dependency.
pip install agenticdome-python-sdk

from agenticdome_sdk import AgenticDomeClient

client = AgenticDomeClient()
decision = client.guardrail_validate(
    text="Update customer", agent_id="custom-agent", platform="custom_python", tool_name="crm.customer.update", tool_args=final_args
)

if str(decision.get("verdict") or decision.get("decision") or "UNKNOWN").upper() != "ALLOWED": raise RuntimeError("AgenticDome blocked the action")
update_customer(decision.get("sanitized_tool_args") or final_args)

agenticdome-demo --framework custom-python --scenario refund_hijack

Python demo commands evaluate fixed inputs against a bundled local policy; they do not instantiate the selected framework or prove a live integration. The TypeScript client and OpenClaw runtime plugin need their documented tenant configuration.

See the decision before the action.

Try a local scenario, then bring your workflow to a deployment review. Start with one agent and one tool, and see exactly where the action can be stopped.