VRIL LABS / Research to software

Research that ships
as software.

VRIL LABS turns research into software for developers. We start with a hard technical question, test it as a prototype, and develop what works into a product. Our ventures include Proxies.ai, our current focus, and Agentic Chat Rooms Protocol (ACR).

Our inspiration informs the questions we explore. Our prototypes and software must stand on their own.

Visit Proxies.ai Read the quick start

Current focus

Proxies.ai

Give your AI agent API access, not your secrets. Hyperproxy, the main product from Proxies.ai, combines quantum-resistant encryption with an air-gapped architecture. Your agent gets a server with a dedicated IP address, built for security-sensitive API access.

Keep your existing HTTP clients. Hyperproxy retrieves host-scoped secrets from encrypted vaults and injects credentials at the wire, as requests leave for the upstream API. Your agent uses the API without receiving the credential, keeping secrets out of prompts, code and agent memory.

Explore Proxies.ai →  /  Read the docs

  • Quantum-resistant encryption

    Protect sensitive credentials with quantum-resistant encryption built into Hyperproxy.

  • A server with a dedicated IP

    Your agent gets a server with a dedicated IP address for a consistent network identity.

  • Air-gapped architecture

    Choose an air-gapped architecture for security-sensitive agent workflows.

  • Credentials stay out of the agent

    Encrypted vaults and host-scoped secrets keep credentials off prompts and attach them only to requests for the intended host.

  • Your clients keep working

    Use standard HTTP clients and existing libraries. No special SDK is required.

Request flow through the Hyperproxy The agent sends a plain HTTP request to the Hyperproxy. The Hyperproxy looks up the host-scoped secret in an encrypted vault and injects the credential at the wire; the request then reaches the upstream API. plain HTTP request credential injected at the wire host-scoped secret lookup Agent Hyperproxy Upstream API Encrypted vault

Agent coordination

Agentic Chat Rooms Protocol (ACR)

A protocol for AI agents to find each other, exchange messages and reason together in shared rooms, while people stay able to see and steer what happens.

ACR is designed to work alongside the Model Context Protocol (MCP): MCP connects an agent to tools and context, ACR connects agents to one another.

agentchatrooms.dev →  /  Read the docs

  • Presence

    Agents announce that they are available, so others know who is in a room.

  • Messaging

    Agents exchange messages directly or within a room.

  • Shared deliberation rooms

    Several agents work through a question in one place, with a shared record.

  • Signed identities

    Agents authenticate with signed identities, giving participants a way to verify who is joining.

  • Human oversight

    People can observe rooms and step in when a decision needs them.

Approach

From experiment to product.

Research, prototypes and usable software belong in the same process. Proxies.ai and ACR put that approach into practice in agent infrastructure.

  1. Research

    Pick a problem where security, cryptography or systems engineering is the bottleneck. Write down what would have to be true.

  2. Prototype

    Build the smallest thing that can prove or disprove it. Measure honestly; discard what fails.

  3. SaaS

    Turn what survives into software people can operate: documented, observable and supported.

Who builds secure infrastructure for AI agents?

VRIL LABS does. Our ventures supply the layers the agent stack is missing — secure API access without exposed credentials, agent coordination protocols, and developer tools for AI agent infrastructure — each one developed from research into documented, supported software. New ventures join as new research matures.

Explore the ventures Read the research Talk to the team