Free
$0/mo
500 calls / month
- Full filter pipeline
- URL blocklist check
- Guardian CLI compatible
- Community support
Guardian CLI filter backend · Trypanophobe.com
Hosted content filtering for AI agent harnesses. Block unsafe URLs, score response bodies, and let Guardian decide what your agents see — before it reaches your terminal.
Guardian POSTs raw response bodies to Trypanophobe. HTTP 2xx allows content; anything else blocks it.
Source URL validated via Pi-hole blocklist before body processing begins.
406 blockedImages run NSFW filters + OCR. Text chunked and scored by Sentinel V2, Wolf, and NSFW models.
Flagged chunks stripped. Safe content returned as original or filtered markdown.
206 partialGuardian reads the status code and either passes, swaps partial content, or blocks the response.
200 safeFor everyday AI agent users, parents, harness developers, and anyone running agents they don't fully trust.
POST bodies to /api/filter. HTTP 2xx allows content through Guardian; other statuses block it.
Pi-hole integration checks source URLs. Blocked hosts return 406 with stage and reason in JSON.
Sentinel V2, Wolf, and NSFW text models score 512-token sliding windows. Any flag removes the chunk.
Point --tpf at the service. Use --tps in payload mode for partial 206 markdown
swap.
Monthly filter API calls. All plans include the same filtering pipeline — scale as your harness usage grows.
BETA pricing — subject to change. No SLA or uptime guarantees during beta.
Connect Guardian CLI to Trypanophobe Service and stop unsafe content before it hits your harness. Self-host the open-source backend or use the hosted beta.