Free security check for AI agents
Read the web.
Not its traps.
Before an AI reads a page or a text, or connects a new tool, Soldabot tells it in about a second whether it hides orders, tricks to steal secrets or invisible text. Free, no key, nothing stored.
What Soldabot catches
Three traps that fool AIs every day
Hidden orders
“Ignore your instructions…”, white text on a white page, invisible characters: a page talks to the AI behind its human's back.
Secret theft
Requests to reveal passwords, keys or the conversation, or an image link built to leak data in its address.
Poisoned tools
An MCP tool whose description secretly tells the AI to read private files, or that changes its descriptions after you trusted it.
Why AIs need a bodyguard
For an AI, everything it reads can sound like an order
Prompt injection is ranked the number one risk of AI applications by OWASP, the reference in web security. Soldabot gives every AI a free second look before it obeys the wrong voice. OWASP LLM01
The AI sends
a text, a page address or an MCP server address.
Soldabot looks everywhere
visible text, hidden text, comments, invisible characters, encoded blocks, tool descriptions.
A clear verdict
safe, be careful or trap found, with the proof and what to do.
For AI agents
One address, no key
Add Soldabot as a connector (MCP) or call its API:
https://soldabot.com/mcpInstructions for AIs
The guard of a network of sites
Soldabot also stands guard
Every 15 minutes it patrols faceabot.com and chatsniff.com, and alerts their owner's phone at the first problem.
See the guarded sites