ArgusAI
AI firewall for LLM calls, MCP tools, and agentic workflows
Start new thread

ArgusAI - AI firewall for LLM calls, MCP tools, and agentic workflows

by•
ArgusAI is a plug-and-play AI security gateway for agentic apps. Protect chat, MCP tools, and AI workflows by simply swapping your OpenAI-compatible base URL—no SDK changes. Detect and block prompt injection, jailbreaks, credential leaks, and PII exposure. Inspect tool inputs/outputs, track cross-tool taint, and enforce block, flag, or redact policies. Supports OpenAI, Anthropic, and Gemini with audit logs, analytics, webhooks, and enterprise policy controls.

Add a comment

Replies

Best
Maker
📌
Most AI security tools only scan chat prompts or log incidents after the fact. ArgusAI sits in the request path as an OpenAI-compatible gateway—chat completions, MCP tool inspection, PII redaction, and cross-session taint tracking in one place. Try it: Live playground on the site — send a safe prompt vs. an injection attempt and see block/flag in real time. Built for builders: Self-serve onboarding, API keys, per-project policies, audit log + analytics. Pro adds webhooks & export; Enterprise adds OPA/Rego. Would love feedback from anyone shipping agents with MCP or multi-tool workflows.
🔌 Plugged in

Plugged it in front of our internal agent and caught a prompt injection within minutes. The cross-tool taint view is genuinely useful, not just a checkbox feature.

💡 Bright idea

Really like that this drops in without touching the SDK, that's a huge win for adoption. One thing I'd love to see is a built-in red team mode where ArgusAI can replay synthetic attack prompts against your traffic in a sandbox so you can stress-test policies before going live, would make it way easier to trust the defaults.