Patronus plans prompt injection detection API: 3MB/s scanning, 1k free requests/day

PatronusProtect · reddit · 2026-08-27

Ben from Patronus vents about existing prompt injection / LLM security APIs: gated behind Contact Sales, per-token pricing, tiny limits, and free tiers you can't actually build on — which he calls a poor model for a security API.

Their plan: a simple API for text, files and URLs, CPU-only, currently hitting 3 MB/s on 6 vCPUs, making large-file scanning cheap; possibly 1,000 free requests/day so developers never hesitate before letting an agent read something. He asks the community what they'd actually need — injection yes/no plus score, exact findings, sanitized content, other threat categories — and whether MCP/WebMCP integration matters for agents or a plain API call suffices.

Original post →

More from coding & agent

coding & agent channel →