4-Person Team Open-Sources Multica to Own the AI Collaboration Layer

A scrappy 4-person team has open-sourced Multica, a framework designed to serve as the collaboration layer for AI agents, timing its release to coincide with Anthropic's latest multi-agent developments. The move signals a growing race to own the critical infrastructure that enables AI agents to work together — a layer that could become as foundational as APIs were to the web era.

Key Takeaways

• A 4-person startup team open-sourced Multica, targeting the 'AI collaboration layer' between agents
• The release was strategically timed alongside Anthropic's own multi-agent announcements
• The founder, a Chinese AI entrepreneur, has pivoted multiple times before settling on this infrastructure bet
• The project bets on a future where the real value lies not in individual agents, but in how they coordinate
• Multica positions itself outside the crowded 'build another agent' space, targeting interoperability instead
• The open-source approach aims to establish a standard before larger players lock down the collaboration protocol

Why the AI Collaboration Layer Matters Now

The AI industry is rapidly shifting from single-model interactions to multi-agent architectures. Companies like Anthropic, OpenAI, and Google have all signaled that the future involves multiple specialized AI agents working in concert — one handling research, another managing code, a third coordinating workflows.

But a critical problem remains unsolved: how do these agents actually talk to each other? Right now, most multi-agent setups rely on ad-hoc integrations, proprietary orchestration frameworks, or brittle prompt-chaining mechanisms. There is no universally adopted protocol for agent-to-agent collaboration.

This gap is exactly what Multica targets. Rather than building yet another AI agent, the team is betting that the real bottleneck — and the real opportunity — lies in the connective tissue between agents. Think of it as the HTTP of the agentic era: a standardized way for AI systems to discover, communicate with, and delegate tasks to one another.

A Serial Pivot Founder Places a Contrarian Bet

The founder behind Multica is no stranger to reinvention. Having pivoted through multiple AI product ideas — a pattern increasingly common among Chinese AI entrepreneurs navigating one of the world's most competitive tech landscapes — the team ultimately landed on infrastructure rather than applications.

This is a deliberately contrarian move. The vast majority of AI startups today are racing to build agents: coding assistants, customer service bots, autonomous research tools. The 'build an agent' playbook has attracted billions in venture capital globally, with companies like Cognition ($175M Series A for Devin), Sierra AI (backed by Bret Taylor), and dozens of Y Combinator startups all competing in the same arena.

By contrast, the Multica team argues that the agent layer is becoming commoditized. As foundation models from Anthropic, OpenAI, and open-source alternatives like Meta's Llama continue to improve, spinning up a capable agent becomes increasingly straightforward. The harder, more defensible problem is making agents interoperate — and that is where Multica aims to plant its flag.

Timing the Release With Anthropic's Moves

The strategic timing of Multica's open-source release is noteworthy. Anthropic has been aggressively expanding its ecosystem play, most notably with the Model Context Protocol (MCP), which provides a standardized way for AI models to connect with external data sources and tools. MCP has gained significant traction, with adoption from companies including Block, Replit, and Sourcegraph.

By releasing Multica alongside Anthropic's latest announcements, the team accomplishes several things:

• Rides the attention wave generated by Anthropic's substantial marketing reach
• Positions Multica as complementary to MCP rather than competitive — MCP connects agents to tools, while Multica connects agents to each other
• Signals urgency to the developer community that the collaboration layer needs an open standard before proprietary solutions dominate
• Demonstrates relevance by showing the project addresses the same architectural challenges Anthropic is tackling

This 'draft behind the leader' strategy is well-established in open source. Projects like LangChain and LlamaIndex similarly gained traction by releasing alongside major model announcements, capturing developer attention at peak moments of industry interest.

What Multica Actually Does

While full technical details are still emerging from the open-source repository, the core premise of Multica centers on providing a standardized protocol and runtime for multi-agent collaboration. Based on the project's positioning, key capabilities likely include:

• Agent discovery: Allowing agents to find and identify other agents with specific capabilities
• Task delegation: Structured protocols for one agent to request work from another
• State synchronization: Keeping multiple agents aligned on shared context and progress
• Conflict resolution: Handling situations where agents produce contradictory outputs or compete for resources
• Observability: Providing visibility into multi-agent workflows for debugging and optimization

The open-source approach is critical here. For a collaboration protocol to succeed, it needs broad adoption — and broad adoption requires trust, transparency, and community ownership. A proprietary collaboration layer would face the same resistance that proprietary messaging protocols encountered in the early internet era.

Industry Context: The Multi-Agent Infrastructure Race

Multica enters a space that is rapidly heating up. Several major players and startups are staking claims in multi-agent infrastructure:

Microsoft has invested heavily in AutoGen, a framework for building multi-agent conversational systems. Google DeepMind has published research on agent collaboration protocols. CrewAI, another open-source framework, has gained popularity for orchestrating role-based agent teams. And Anthropic's own work on MCP represents a foundational layer that multi-agent systems will likely build upon.

Compared to frameworks like AutoGen or CrewAI, Multica appears to focus more narrowly on the communication and coordination protocol rather than the full orchestration stack. This is a strategic distinction — by not trying to be a complete agent framework, Multica can potentially serve as a neutral layer that works across different agent implementations.

The total addressable market for AI agent infrastructure is substantial. According to recent estimates from venture capital firms, the agentic AI market could reach $50 billion by 2028, with infrastructure and middleware capturing a significant share of that value.

What This Means for Developers and Businesses

For developers building multi-agent systems today, Multica's open-source release offers a potential standard to rally around. The current landscape is fragmented — most teams building multi-agent applications are rolling their own coordination logic, leading to duplicated effort and incompatible systems.

For businesses evaluating agentic AI deployments, the emergence of collaboration protocols like Multica suggests that multi-agent architectures are maturing beyond research prototypes. Companies planning AI strategies should consider how their agents will interact not just with tools and data, but with other agents — both internal and external.

The key question is whether a 4-person team can build enough momentum to establish a standard in a space where well-funded competitors are actively operating. Open-source dynamics could work in their favor: if the protocol is well-designed and solves a genuine pain point, community adoption can outpace even the largest corporate efforts.

Looking Ahead: Can a Tiny Team Set the Standard?

The history of internet infrastructure offers both encouragement and caution for the Multica team. Many foundational protocols were created by small groups — HTTP, SMTP, and RSS all emerged from modest origins. But the modern AI landscape moves faster and involves players with far deeper pockets.

Several factors will determine Multica's trajectory over the next 6 to 12 months:

• Developer adoption: Early GitHub stars and community contributions will signal viability
• Integration partnerships: Getting major agent frameworks to support Multica natively
• Anthropic's own moves: Whether Anthropic extends MCP into agent-to-agent territory, potentially competing directly
• Funding: Whether the team secures venture backing to scale beyond 4 people
• Standardization efforts: Whether industry bodies or consortia form around multi-agent protocols

The founder's history of multiple pivots could be either a strength or a liability. On one hand, it demonstrates adaptability and willingness to find product-market fit. On the other, investors and potential adopters may question the team's long-term commitment to this specific vision.

What is clear is that the AI collaboration layer represents a genuine gap in the current stack. Whether Multica fills that gap or merely highlights it for larger players to address, the project underscores a critical truth: the future of AI is not about single agents acting alone, but about ecosystems of agents working together. The team that defines how that collaboration happens could own one of the most valuable layers in the entire AI stack.