eth-agents: An Ethereum Multi-Agent Development Platform Based on Claude Code

eth-agents is an Ethereum development plugin designed specifically for Claude Code. It achieves end-to-end automation of smart contract development, security auditing, testing, gas optimization, and multi-chain deployment through collaboration among 10 professional AI agents. Developers can complete complex tasks simply by describing their requirements in natural language, without the need to frequently switch tools or memorize command-line parameters.

Traditional Ethereum development requires switching between tools like Foundry and Hardhat, which consumes effort in configuring environments and scripts. eth-agents adopts zero configuration, natural language interface, and specialized division of labor as its core design principles, encapsulating complexity under the agent collaboration layer. It is suitable for scenarios such as rapid prototyping, security auditing, multi-chain deployment, and team collaboration.

The system centers on the Orchestrator (orchestrator), which coordinates 10 agents:

• Orchestration Layer: The Orchestrator is responsible for task scheduling, model selection, state updates, and result verification;
• Development Layer: Contract Developer (contract writing), Unit Test Engineer (unit testing), Integration Test Engineer (integration testing);
• Security Layer: Auditor (auditing), Security Specialist (vulnerability analysis);
• Operations Layer: Gas Optimizer (gas optimization), Deploy Engineer (multi-chain deployment), dApp Developer (frontend synchronization);
• Debugging Layer: Debugger (root cause analysis).

eth-agents encapsulates nine core capabilities, corresponding to the agent pipeline:

1. audit: Security vulnerability analysis (including static/dynamic/formal verification);
2. tdd: Test-driven development (write failing tests first, then implement code);
3. build-contract: End-to-end contract construction (requirements → TDD → auditing → gas optimization);
4. deploy: Multi-chain deployment (dry-run preview, user confirmation, frontend update);
5. gas-optimize: Gas optimization (improvements must be verified via forge snapshot);
6. fork-test: Mainnet fork testing (automatic debugging on failure);
7. full-protocol: End-to-end protocol construction (multi-stage pipeline);
8. learn: Save reusable workflow patterns;
9. help: Display capability guides.

Installation Methods:

• GitHub Installation: claude plugin install github:cayocan/eth-agents;
• Community Repository Installation: First add the community repository, then install eth-agents@claude-community. Interaction Examples:
• Auditing: "Perform a complete security audit including formal verification";
• Development: "Build an ERC4626 staking contract";
• Deployment: "Deploy to Arbitrum Sepolia and update the frontend";
• Full Workflow: "Build a complete DeFi protocol from scratch".

Security Protection:

1. Dry-run Preview: Simulate execution before deployment to show plans and costs;
2. Explicit Confirmation: Mainnet deployment requires users to input "confirm";
3. Pre-Auditing: build-contract and full-protocol automatically perform static audits;
4. Vulnerability Escalation: Critical/High vulnerabilities trigger in-depth analysis by security specialists. State Visualization: Integrates the MCP HUD server to display real-time capability activation, progress, token usage, and agent status.

Application Scenarios:

• Rapid Prototyping: Quickly generate runnable contracts using natural language;
• Security-First: Audit processes covering static, dynamic, and formal verification;
• Multi-Chain Strategy: One-click deployment to multiple networks with automatic address management;
• Team Standardization: Precipitate best practices to ensure consistent workflows;
• Learning and Growth: Newcomers learn professional processes by observing agent collaboration. Value: Simplifies complex development into natural language conversations, lowering the barrier to Ethereum development while maintaining professional-level quality and security.