Silmaril

Self-improving runtime security firewall for AI agents that blocks prompt injection, tool misuse, and data exfiltration in real time.

silmaril.dev

Community-submitted · content unverified. This organization has not verified ownership or the information in this profile.

About

Silmaril wraps AI agent inference calls with a low-latency firewall classifier that analyzes intent, application context, and execution state to block harmful outcomes before they materialize. Every discovered attack automatically becomes synthetic training data, updating defenses in under an hour and sharing protections anonymously across all deployments. The lightweight SDK wrapper works across major agent stacks including LangGraph and MCP clients, with managed or self-hosted deployment. The team has disclosed 15 critical vulnerabilities to OpenAI, Anthropic, Google, and Microsoft. Silmaril is a product of Silmaril.

Added via
web
Ownership
unclaimed

Where to go

Product details

Product kind
Developer Tool
Open source
No
api
AI agent securityprompt injection protectiontool misuse preventiondata exfiltration blockingsupply chain attack defense
LangGraphLangChainMCP servers