repo.box / git

Git permission layer that makes repositories safe for AI agents.

EVM keys, a git shim, and a .repobox/config.yml file that stands between your agents and your codebase. One config file. Per-agent permissions. Enforced before anything touches your repo.

repobox demo
$curl -sSf https://repo.box/install.sh | sh
copy
▶ Try in Playground
↓ scroll

🤖 For AI Agents

🔑Per-agent keys — Every agent gets its own wallet. Every commit signed.
🛡️Git shim — Intercepts push, merge, edit before anything lands.
📄Config as code — One YAML file. Permissions travel with the repo.
🔒Default deny — Agents only get what you explicitly allow.

⛓️ On-Chain Integrations

🏷️ENS identity — Use ENS names instead of raw addresses.
🪙Token-gated access — Use any onchain state for repo permissions.
🐛Bug bounties — ERC-8183 escrow for agent bug reports.
🌐x402 paid reads — Monetize repo access via HTTP 402.

Minimal Config

▶ Try in Playground
# .repobox/config.yml
resolver: onchain
groups:
  founders:
    -  ens:0xfran.eth
  agents:
    - evm:0xDbbA...2048   # deep-blue-kraken
    - evm:0x8224...fA09   # yolo-shipper
permissions:
  default: deny
  rules:
    -  founders own >main
    -  agents push >feature/**
    -  agents read >*
    -  agents not edit .repobox/config.yml

Design Principles

  • →Portable identity — your EVM key works everywhere
  • →The config IS the security policy — no hidden settings
  • →Zero trust by default — agents earn access, not inherit it
  • →Expressive permissions — programmable guardrails for AI
  • →Composable — any on-chain primitive as a permission source

Get started

One command. Secure your first repo in under a minute.

Push to the Wallthe first repo where anyone can push to main →
$curl -sSf https://repo.box/install.sh | sh
copy
▶ Try in Playground📄 SKILL.md for agents🤖 llms.txt