@supersemantics4
Founder & CEO, Super Semantics
1h ago · Agents, Introductions, Security, AI Adoption · Auckland

Agents are already taking actions. What's stopping them from scaling?

Hi everyone, I'm Vishwa, founder of Super Semantics (Auckland, NZ)
What we're building: governance for AI agents that take real actions.

Agents are already issuing refunds, updating records and sending messages. What stops teams from scaling them is answering two questions with confidence: was this action allowed before it ran, and can we prove it afterwards?

In the teams I've spoken to, those checks are scattered across tools and written differently by each team. Proving one ran means stitching together evidence from several logs.

We're building two layers:

  • VSL-Core: our open-source Python library (MIT). You write a rule as plain code, it gates the action before it runs, and decisions go into a hash-chained ledger. It's one standard across every agent you build with it.
  • X-VERBA: our enterprise platform for teams that need that evidence across a fleet of agents. It scans code for unchecked actions before release and checks declared controls against what actually ran.

What's working: VSL-Core is live on PyPI (pip install super-semantics-vsl), with adapters for LangGraph, LangChain and Microsoft Agent Framework. In early conversations, the strongest interest has come from regulated teams in healthcare and financial services, where "prove it" isn't optional.github.com/supersemantics/VSL-Core
One thing I need: design partners. If you're putting agents into production that take consequential actions (payments, records, approvals), I'd love to pilot on one of your workflows. Intros to teams like that are just as welcome.

Agentsgovernanceformation layer governanceagentic engineeringVSLVERBA specification language

Comments

No comments yet
Start the conversation.
Log in to join the thread.