Introducing Checkrd

AI agents are making real API calls with real consequences. They need infrastructure-level controls, not just prompt engineering.

The Problem

When an AI agent calls an external API, there is no policy enforcement at the network layer. There is no kill switch. There is no audit trail.

These are not AI problems. These are infrastructure problems. The agents work. The rails they run on do not exist.

Our Approach

Checkrd is a proxy layer that sits between your agent and every external API it calls. It provides:

• Policy enforcement -- define which APIs, methods, and endpoints your agent can access
• Kill switch -- instantly halt all outbound traffic for any agent
• Observability -- structured telemetry for every request, with OpenTelemetry compatibility

How It Works

A 233KB WebAssembly core handles policy evaluation in 62 microseconds. Language wrappers load the WASM module and patch your HTTP client transparently.

from checkrd import wrap
import httpx

client = wrap(httpx.Client(), agent_id="sales-agent", policy="policy.yaml")
# Every request through this client is now policy-enforced

The control plane provides a dashboard, real-time telemetry, and team management.

Open Source

The proxy core and language wrappers are open source. The control plane is hosted SaaS. The boundary is clean: anything a solo developer needs is free. Anything a team needs is paid.