SOLUTIONS / AGENT BUILDERS

Protect your agents

Drop-in prompt injection protection for MCP and A2A agents. Layered latency — ~50µs deterministic rules, <2ms typical end-to-end — means your agents stay fast while staying safe.

View pricing Read the docs

Technical specifications

<2ms

Latency

typical · deterministic path ~50µs

1.27%

False Positive Rate

industry-leading accuracy

MCP + A2A

Protocols

native support

90 days

Evidence Retention

audit trail included

Integration

zn is an HTTP API you can call from anywhere — any language, any agent framework. A Python client is bundled with the source release. Add zn to your agent in minutes:

bashcurl -X POST https://api.usezn.com/prod/analyze \
  -H "Authorization: Bearer $ZN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"input": "the tool-call input your agent is about to execute"}'

# verdict "block" -> reject the input before the tool call
# verdict "allow" -> safe to proceed

Features

Real-time prompt injection detection

Deterministic, versioned rules engine

Evidence vault with SHA-256 hashing

Structured threat classification

Zero-config MCP integration

A2A protocol support

Webhook notifications

Rate limiting and quotas

How it works

1. Intercept

Every tool call passes through zn before reaching your agent. We analyze the input in real-time.

2. Analyze

Our detection engine identifies prompt injection patterns, instruction overrides, and data exfiltration attempts.

3. Protect

Threats are blocked, evidence is logged, and your agent continues operating safely.

Ready to protect your agents?

Start free with a 48-hour Hobby trial. No credit card required.

View pricing