BAND, an Israeli startup also known as Thenvoi AI Ltd., launched from stealth on April 23 with $17 million in seed funding to build communication infrastructure for multi-agent AI systems. The round was led by Sierra Ventures, Hetz Ventures, and Team8.
The company is targeting what it describes as the fragmentation problem in enterprise agent deployments: agents built on LangChain cannot hand off tasks to those built on CrewAI, and a Salesforce-embedded agent has no native way to coordinate with a custom Python script running on a private cloud. BAND’s platform provides a framework-agnostic interaction layer that lets agents discover each other, exchange context, delegate tasks, and collaborate in real time, according to VentureBeat.
How the Agentic Mesh Works
BAND’s architecture splits into two layers. The first is a communication mesh where agent discovery and structured delegation occur across clouds and frameworks without requiring developers to write custom integration code for each connection.
The second is a control plane that provides runtime governance: authority boundaries that restrict which agents can communicate, credential traversal that manages how human permissions flow through agent delegation chains, and monitoring for task handoffs.
One technical decision stands out. BAND does not use LLMs to route messages between agents. “Using an LLM for routing would introduce the same non-deterministic errors the platform seeks to solve,” the company told VentureBeat. Instead, the platform uses a patent-pending deterministic routing architecture. The underlying infrastructure is built on messaging stacks comparable to WhatsApp and Discord, designed to handle billions of agent-to-agent messages at scale.
The platform supports full-duplex, multi-peer communication. Unlike existing protocols that are primarily peer-to-peer or client-server, BAND allows groups of agents (a planning agent, a coding agent, and a QA agent, for example) to work together in a shared “room” with synchronized context.
Positioning Against Hyperscalers
BAND enters a market where the largest model providers are building their own orchestration layers. OpenAI announced workspace agents on April 22. Anthropic launched Claude Managed Agents earlier this month. Google’s Gemini Enterprise Agent Platform debuted at Cloud Next 2026.
CEO Arick Goomanovsky told VentureBeat that BAND “plays the role of the independent platform that allows an enterprise to avoid vendor lock-in.” The pitch: enterprises running agents across multiple model providers need a neutral coordination layer, not one tied to a specific vendor’s ecosystem.
“We’re entering the agentic economy, where millions of agents will need to collaborate across companies, platforms and environments,” Goomanovsky said in the company’s press release. “The challenge isn’t only building more agents, but getting them to work together in real time.”
Tim Guleri, managing director at Sierra Ventures, framed the investment around infrastructure timing. “Multi-agent systems are quickly becoming the foundation of modern software,” Guleri told Ynet. “Without a reliable and efficient way for agents to communicate, their potential is limited.”
The Coordination Gap in Practice
The company said its platform supports agents built with LangChain, CrewAI, and custom frameworks, as well as SaaS-embedded agents and personal AI assistants including OpenClaw, Claude Code, and Codex. Governance tools provide visibility into agent interactions, monitor task delegation, and enforce authority boundaries.
BAND’s bet is that coordination, not capability, becomes the bottleneck as enterprises scale from dozens to hundreds of autonomous agents across engineering, security, and operations workflows. That’s a bet shared by its investors: $17 million at seed stage for infrastructure middleware reflects conviction that the orchestration layer is where value accrues next.