Backslash Security

Backslash Security is an Agentic AI Endpoint Security platform. It secures the agentic AI fabric running on enterprise employee endpoints: AI agents, the models they call, MCP servers and their individual tools, Skills, hooks, rules files and plugins. That layer executes locally with the user's own permissions, changes frequently, and produces neither a malware signature nor network traffic - placing it outside what EDR, DLP and network or AI gateways were built to inspect. The platform covers five functions: Discover - continuous inventory of every AI agent, model, MCP server, Skill, hook, plugin and local model across the endpoint fleet, including components installed under personal accounts or outside security oversight. Assess - continuous risk analysis of each component and of combinations of components, mapped to OWASP and MITRE ATT&CK, covering prompt injection, hidden instructions, excessive permissions, supply chain risk and configuration drift. Govern - policy enforcement down to the individual MCP tool and Skill, scoped by user group, with policy packs. Protect - real-time verification of agent behaviour against intent, blocking risky actions inline, including an endpoint MCP proxy that inspects agent-to-server communication locally. Investigate - forensic audit trail connecting prompts, loaded Skills, MCP interactions, tool calls and resulting actions, supporting incident response and evidence for EU AI Act, NIS2, DORA and SOC 2. Supported tools include Claude Code, Cursor, GitHub Copilot, OpenAI Codex, Gemini CLI, Google Antigravity, Devin Desktop, OpenClaw, Claude Cowork and any MCP-compatible agent. Deployment is via MDM and designed to coexist with existing EDR. An agentless, read-only assessment is available for initial discovery. Backslash also operates free public tools: an MCP Server Security Hub with 81,000+ public MCP servers assessed, a Skills Security Scanner, and Claw-Hunter (open source).