AgentSeal is an open source security toolkit that protects your entire AI agent stack. Four core commands, one mission: find vulnerabilities before attackers do. ✓・agentseal scan - 191+ adversarial probes testing extraction and injection resistance. Deterministic results, no LLM judges. ✓・agentseal guard - Machine-wide scanning across 17+ AI agent types (Claude, Cursor, Windsurf, Copilot, Cline and more). Detects poisoned skill files, dangerous MCP configs, obfuscated payloads. ✓・agentseal shield - Real time background monitoring with desktop alerts. Like antivirus for your agent stack. ✓・agentseal scan-mcp - Runtime MCP server analysis, toxic flow detection, rug pull identification. All runs locally. Privacy first. Pro adds MCP poisoning probes, RAG poisoning, multimodal attacks, behavioral genome mapping, and PDF reports. Docs: agentseal.org/docs GitHub: github.com/AgentSeal/agentseal Install: pip install agentseal
| Website | http://agentseal.org |
| Employees | 1 (0 on RocketReach) |
| Founded | 2026 |
| Industry | Technology, Information and Internet |
Looking for a particular AgentSeal employee's phone or email?
AgentSeal Questions
