agentsploit/agentsploit

AgentSploit

Offensive security framework for AI agents and MCP servers.

AgentSploit is a Burp Suite / Metasploit-style framework purpose-built for the agentic AI attack surface. It helps red teamers, AI security researchers, and product security teams probe LLM agents and Model Context Protocol (MCP) servers for vulnerabilities that legacy tooling cannot find.

[!IMPORTANT] AgentSploit is an authorized-use security testing tool. You must have explicit written permission to scan any target you do not own. See AUTHORIZATION.md before running anything.

👉 New here? Start with docs/getting-started.md - a 10-minute tour of every capability against bundled fixtures, no API keys needed.

---

Why this exists

Every Fortune 500 is shipping LLM agents and MCP servers in 2026. The attack surface is genuinely new:

• Tool descriptions are LLM-readable instructions - malicious ones can hijack agents.
• Agents fetch untrusted content from PDFs, web pages, calendar invites, tickets - and that content can issue commands.
• Chained tool calls create privilege escalation paths that no traditional permission model captures.
• Memory and context windows can be poisoned across sessions.

Existing scanners (Burp, ZAP, Semgrep, Snyk) don't speak this layer. AgentSploit does.

TL;DR

pip install agentsploit

# Scaffold an engagement
agentsploit init my-engagement/ --authorized-by "Jane D