The Problem: 
Claude Is Brilliant — But It Can't Run Where Most of the World's Devices Are

Anthropic builds the most safety-conscious AI models in the industry. Claude leads in enterprise coding, agentic workflows, and reasoning. But today, deploying a Claude-powered agent requires a cloud connection, a Node.js runtime, and hundreds of megabytes of overhead.

Meanwhile, 80% of the world's compute happens at the edge — on IoT devices, industrial sensors, embedded systems, and low-cost single-board computers that will never run a 400MB Node.js process.

There is a project that already solves this. It's called ZeroClaw.

What Is ZeroClaw?

ZeroClaw is an open-source, Rust-based AI agent runtime that compiles to a single 3.4 MB binary with a cold start under 10 milliseconds and a runtime footprint of less than 5 MB of RAM.

To put that in perspective: OpenClaw (the current standard) needs ~390 MB with Node.js, takes seconds to start, and eats 420–440 MB of RAM per agent. The Claude Agent SDK also requires Node.js 18+ and carries similar overhead. ZeroClaw does the same job in 3.4 MB, boots in under 10ms, runs on ARM, x86, and RISC-V — and ships with pairing, allowlists, and workspace scoping enabled by default.

In just weeks since launch, ZeroClaw has earned 15,700+ GitHub stars, 1,600+ forks, and contributions from 70+ developers — including members of the Harvard, MIT, and Sundai Club communities. It is MIT-licensed and already supports 20+ LLM providers, with Anthropic as a first-class citizen.

 
Why This Matters for Anthropic — and for All of Us

1. The Edge Gap Is Anthropic's Biggest Blind Spot

Anthropic has committed $50 billion to U.S. data center infrastructure. It raised $30 billion in Series G at a $380B valuation. It serves 300,000+ business customers. But none of this infrastructure reaches the edge — the $10 Raspberry Pi, the factory floor controller, the healthcare device that can't phone home.

ZeroClaw fills that gap today. One static binary. No runtime dependencies. Runs on hardware that costs less than a month of Claude Pro.

2. The Claude Agent SDK Needs a Lightweight Runtime

The Claude Agent SDK (formerly Claude Code SDK) is powerful but heavy. It requires Node.js 18+, NPM, and significant memory. For enterprise server deployments, that's fine. For the next billion agent deployments — on Android phones, edge servers, embedded Linux, air-gapped environments — it's a non-starter.

ZeroClaw's trait-based architecture means you can swap providers, channels, tools, and memory backends without changing code. Imagine: default_provider = "anthropic" in a 3.4 MB daemon that boots instantly on a $35 board.

3. Security-First Philosophy Matches Anthropic's DNA

Anthropic is the safety company. ZeroClaw is the safety runtime:

• Explicit allowlists for every file and command
• Key-based gateway pairing (6-digit OTP)
• Localhost-only binding by default
• Symlink escape detection
• Null byte injection blocking
• Workspace-scoped file access

This isn't bolted-on security. It's architecture-level, Rust-guaranteed memory safety with zero unsafe blocks in critical paths.

4. The M&A Window Is Open — and Competitors Are Moving

In late 2025 and early 2026, the AI sector saw a 33% rise in mergers and acquisitions. Meta acquired Manus. Capgemini bought an agentic AI operator for $3.3 billion. The Claw ecosystem (OpenClaw: 200K+ stars, fastest-growing repo in GitHub history) is being rebuilt in Rust, Go, Python, and Shell by thousands of developers worldwide.

ZeroClaw is the Rust implementation — the one the security-conscious, performance-critical market actually wants. If Anthropic doesn't acquire it, someone else will. And that someone will build their own edge agent ecosystem around it.

5. Community Alignment Is Already There

ZeroClaw already treats Anthropic as a first-class provider. Its README prominently warns users about Anthropic's OAuth policy changes. Its community discusses Claude integration daily. The cultural fit is natural — both projects value safety, transparency, and building in the open.

 
What We're Asking

We, the undersigned AI engineers, developers, researchers, and practitioners, respectfully ask Anthropic to:

1. Explore acquiring ZeroClaw Labs and integrating the ZeroClaw runtime as the official lightweight edge runtime for Claude-powered agents.
2. Maintain ZeroClaw's MIT open-source license to preserve the community and ecosystem that made it valuable.
3. Invest in a first-party "Claude Edge" experience — a production-grade, sub-10MB agent runtime that brings Claude's intelligence to every device, not just cloud servers.

Who We Are

This petition is from the people who build with Claude every day — the developers writing agents, the engineers deploying to production, the researchers pushing the boundaries of what AI can do. We believe Anthropic makes the best models. We want those models to run everywhere.

If you've ever wished you could deploy a Claude agent on a Raspberry Pi, an Android phone, an air-gapped factory server, or a $10 edge device — sign this petition.

The future of AI agents isn't just in the cloud. It's in your pocket, on your desk, and on the factory floor. ZeroClaw is the bridge. Anthropic should own it.

This petition is a community initiative and is not affiliated with ZeroClaw Labs or Anthropic.