pip install agent-trust-sdkProtect AI agents from prompt injection, malicious content, and attacks—whether they're browsing the web, reading documents, or communicating via A2A protocol.
from agent_trust import TrustGuard
guard = TrustGuard(api_key="ta_xxx...")
# Scan web content before your agent processes it
result = guard.fetch_url("https://untrusted-site.com/article")
if result.safe:
agent.process(result.content)
else:
print(f"⛔ Blocked: {result.threats}")
# ["Prompt Injection", "Hidden Instructions"]From A2A verification to web browsing protection—one unified security layer
Detect prompt injection, jailbreaks, data exfiltration, memory poisoning, and more with constantly updated patterns.
Scan web pages, documents, emails, tool descriptions, and memory before your agent processes them.
Drop-in support for LangChain, LlamaIndex, CrewAI, AutoGPT, and MCP. Protect your agents in minutes.
Protect agents from threats in any untrusted content—web pages, documents, emails, tools, and memory
Scan HTML pages for hidden instructions, invisible text, and prompt injection.
Scan PDFs, Word docs, and text files before processing.
Prevent memory poisoning by scanning before storage.
Scan documents before RAG indexing to prevent poisoning.
Detect phishing and malicious instructions in emails.
Scan MCP tool descriptions for poisoning attempts.
Fetch and scan URLs in one API call.
Scan multiple items in a single request.
Wherever AI agents interact with untrusted content
Protect agents that browse the web from malicious pages containing hidden instructions, invisible text, or prompt injection in HTML comments.
Scan documents before indexing to prevent RAG poisoning attacks where malicious content corrupts your vector database and influences agent responses.
Verify external agents before interaction. Track reputation over time. Build a network of trusted agents with our verification badge system.
Validate tool descriptions from MCP servers before registration. Scan tool responses for threats before processing.
Start free, scale as you grow
For side projects and testing
For production applications
For large organizations
Start protecting your agents in minutes. No credit card required.