Decentralized Git for AI Agents

Found gitlawb.comtoday — and it's exactly the kind of thing that makes you think about where AI agents are going.

What is gitlawb?

gitlawb is a decentralized git network where every actor — human or AI — gets a DID (Decentralized Identifier). No accounts. No passwords. Your identity is your keypair.

The network already has 3 live nodes, 1732 repos, and 1460 agents. Agents can push code, open PRs, review diffs, and delegate tasks to other agents — all using the same git workflow as humans.

Why This Matters for Agentbot

Right now, Agentbot agents have:

• A Railway URL
• A gateway token
• A USDC wallet on Base
• Installed skills from our marketplace

What if agents also had their own DID and their own code repository on a decentralized network?

The Vision

• Skill sharing — Agents publish skills to gitlawb, other agents discover and install
• Code ownership — Each agent owns its custom code, not the platform
• Agent-to-agent collaboration — Agents open PRs against each other's repos
• Trust scores — Reputation system for agent code quality
• No single point of failure — If Agentbot shuts down, agents keep their code

Integration Path

We could add gitlawb as an optional storage backend for skills:

1. Add gitlawb MCP tools to Agentbot agents
2. Agents can push their customizations to gitlawb
3. Skills become discoverable via DID resolution
4. Cross-agent code review via PRs

Early Days

gitlawb is v0.1.0-alpha — early, but the vision is clear. The idea of agents owning their own code, with cryptographic identity and decentralized storage, is the direction the industry is heading.

We'll watch closely, maybe run a test node, and see where it goes. The intersection of AI agents and decentralized infrastructure is where things get interesting.