AgentShield for Anthropic

Monitor Claude API usage, track costs per session, detect prompt injection attempts on your Anthropic-powered agents.

Quick setup

import agentshield
import anthropic

shield = agentshield.init(api_key="your-key")
client = shield.wrap_anthropic(anthropic.Anthropic())

# All Claude calls now tracked
message = client.messages.create(
    model="claude-opus-4-5",
    max_tokens=1024,
    messages=[{"role": "user", "content": prompt}]
)