Skip to content

About Amla Labs

Building the security infrastructure for multi-agent AI systems.

Our Mission

We're solving the agent delegation gap—the critical security problem of how AI agents safely share credentials without granting permanent access.

By applying capability-based security principles from operating systems research (seL4, Mach) to AI agent orchestration, we enable enterprises to deploy multi-agent systems without compromising security.

The Problem We're Solving

Phase 1: Credential Storage

Where do enterprises securely store API keys, OAuth tokens, and database credentials that AI agents need?

Current solutions: Environment variables (insecure), secret managers (not agent-aware), or manual copy-paste (worse).

Phase 2: Agent Delegation

KEY DIFFERENTIATOR

How to let Agent A safely delegate credentials to Agent B without granting permanent access?

Our solution: Cryptographic capabilities with send-once enforcement, progressive attenuation, and provable delegation chains.

Phase 3: Auto-Attenuation

How to automatically compute minimal permissions from agent task descriptions?

Future: LLM-powered Agent Card parsing that extracts constraints from natural language task descriptions.

Why Now?

1

Multi-Agent Systems Are Mainstream

46% of YC Winter 2025 batch are building multi-agent products. CrewAI, LangGraph, AutoGen, and Semantic Kernel have made orchestration accessible, but credential management is still manual and insecure.

2

Validated by Security Research

Palo Alto Networks Unit 42 (2025) demonstrated Agent Session Smuggling attacks. Their recommended defense—cryptographically signed agent credentials—validates our approach.

3

Enterprises Need Compliance

SOC 2, PCI DSS, and FedRAMP all require auditable credential access. Current solutions can't prove which agent accessed what credential or whether delegation was authorized.

Founder Background

Amla Labs was founded by engineers with deep experience in capability-based security and formally verified systems.

  • Apple Core OS - Experience building secure operating system primitives for macOS and iOS
  • seL4 Research - Formal methods and capability-based security in microkernels
  • Distributed Systems - Building high-performance, fault-tolerant infrastructure

"I spent years working on capability-based systems for operating systems. When I saw multi-agent AI systems emerging, I immediately recognized the same security problems we solved decades ago. The industry needs cryptographically enforceable boundaries for AI agents."

We're Hiring

Looking for Enterprise Security Sales Co-Founder

We're building a deeply technical product for enterprise security teams. Looking for a co-founder who:

  • Understands enterprise security buyer personas (CISOs, security architects)
  • Can articulate technical value proposition (OAuth vs capabilities)
  • Excited about AI agent security (new market category)
Get in Touch

Design Partner Program

We're looking for 5-10 companies building multi-agent systems to help shape the product. Ideal design partners are:

Building with CrewAI, LangGraph, or AutoGen
Need SOC 2, PCI DSS, or FedRAMP compliance
Orchestrating 3+ agents per workflow
Currently managing credentials manually
Apply to Join

Get in Touch

General Inquiries

[email protected]

Follow Us