Runtime controls for production agents

AgentReady Runtime

Policy packs, decision tracing, and an MCP proxy — enforce and observe your agents at runtime, grounded in the same standards as your report.

Policy packs

Every paid report includes a downloadable policy pack: ten agent-security policies compiled from your answers and findings, each citing the standard and vendor practice it implements. Feed the JSON to the AgentReady MCP proxy for pre-action enforcement on tool calls.

Start a free check → unlock a report

Trace SDK (open schema)

Lightweight stdlib-only Python SDK. Buffers decision rows and POSTs them to /api/trace using your report's access token. The schema is public and mirrors our own evidence tables.

Download SDK View schema.json

from agentready_runtime_trace import AgentReadyTrace
tr = AgentReadyTrace(base_url='https://...', api_key='<report-token>', report_id='rpt_...')
tr.record('checkout-agent', 'tool_call', {'action': 'refund'}, input_ref='sess-123')
tr.flush()

MCP proxy

A stdio proxy that sits in front of any MCP server and evaluates every tools/call against your policy pack before it reaches the real server — block, require approval, or audit. Emits a local audit log and optional trace ingest.

Wrap any server: python agentready_proxy.py --policies pack.json -- npx your-mcp-server

Traces are stored only for paid reports and are accessible to the report owner. Runtime observability for your agents — not a certification.