Claude Code Plugin · Open Source

eth-agents

A team of 10 specialized AI agents for Ethereum smart contract development.
Built for Claude Code. Zero configuration. Natural language interface.

10 Specialized Agents
9 Skills / Pipelines
~55% Token Savings
0 Config Files
Get Started View on GitHub
$ claude plugin install eth-agents
eth-agents mascot

Describe. Delegate. Ship.

You write in plain English. The orchestrator routes your intent to the right agent pipeline. Agents coordinate automatically — no commands to remember, no toolchain configuration.

claude code — eth-agents
you › audit this contract for reentrancy   orchestrator → routing to audit pipeline auditor → static analysis · slither scan HIGH-1: releasePrize — winner not validated HIGH-2: no ReentrancyGuard on external fns → 2 HIGH · 4 MED · 3 LOW · 1 INFO security-specialist → exploit simulation PoC: cross-function reentrancy via callback   audit-report.md written 10 findings · remediation for each   you ›
  1. 01

    Describe in plain English

    No flags, no commands. "Audit this contract", "build a staking contract with ERC4626", "deploy to Arbitrum Sepolia" — eth-agents understands intent.

  2. 02

    Orchestrator routes the pipeline

    A coordinator agent identifies the skill, selects the right agents in the right order, and manages handoffs automatically. You see progress in the Claude Code status bar.

  3. 03

    Specialized agents execute

    Each agent has a specific role, defined tools, and output constraints. They produce structured artifacts — audit reports, test files, gas diffs, deployment receipts.

  4. 04

    You review, never blocked

    Mainnet deployments always halt for your explicit confirmation. All agents report what they did and why. You stay in control.

10 Agents.
One orchestrator.

Each agent is a markdown file with a defined role, allowed tools, output constraints, and handoff rules. Model routing keeps costs low — Haiku for simple tasks, Sonnet for implementation, Opus for security.

eth-agents team — Auditor, Developer, Deploy, Security
auditor developer deploy security + 6 more
Coordinator
Orchestrator
Entry point for every skill invocation. Interprets intent, selects the skill pipeline, spawns the right agents in order, manages token budget, and emits MCP HUD status updates.
model: opus
Development
Contract Developer
Writes production-ready Solidity contracts. Spec → TDD loop → gas pass. Never ships without green tests.
model: sonnet
Security
Auditor
Systematic vulnerability analysis with Slither/Mythril. Configurable depth: static, dynamic (Echidna), or formal (Certora/Halmos).
model: opus
Security
Security Specialist
Activated on Critical/High findings. Writes PoC exploit tests, simulates attack vectors, confirms impact before marking a finding.
model: opus
Testing
Unit Test Engineer
Writes isolated unit tests before implementation exists. Mocks all external deps. Runs regression after implementation.
model: sonnet
Testing
Integration Test Engineer
Fork tests, fuzz tests, and invariant tests against real mainnet state. Always pins block number — never uses `latest`.
model: sonnet
Testing
Debugger
Diagnoses failing transactions with `cast run`, `forge trace`, `forge debug`. Documents root cause before suggesting any fix.
model: sonnet
Operational
Gas Optimizer
Reduces gas without changing behavior. One category at a time. Must produce a positive diff before handing off.
model: sonnet
Operational
Deploy Engineer
Multichain deployment scripts, dry-runs, Etherscan verification. Never deploys to mainnet without your explicit confirmation.
model: sonnet
Operational
dApp Developer
Auto-detects your stack (wagmi/viem, ethers.js, Next.js/React/Vue). Reads ABIs from deployment artifacts — never copy-paste.
model: sonnet

9 Pipelines.
One command each.

Skills are entry points. They activate coordinated agent pipelines. You trigger them by describing intent — no syntax to memorize.

What an audit
looks like.

eth-agents generates structured, actionable audit reports. Every finding includes severity, impact, location, and a concrete remediation with code.

audit-report.md completed
0 Critical
2 High
4 Medium
3 Low
1 Info
HIGH

Winner address not validated in prize release

Privileged function can redirect funds to any arbitrary address. Direct fund loss vector.

HIGH

No reentrancy guard on external functions

Cross-function reentrancy possible via callback-capable token. Risk of double refund.

MED

Missing SafeERC20 — silent failure risk

Raw transfer calls will revert with non-standard ERC-20 tokens, bricking deposit/refund flows.

MED

Single immutable owner — no transfer mechanism

Lost key locks all rooms. Compromised key gives full prize control. No Ownable2Step.

MED

Asymmetric stake amounts — unfair betting risk

No enforcement that both players stake the same amount. Griefing via front-running possible.

LOW

block.timestamp used for 45s stake window

Validator manipulation (~12–15s) represents ~33% of the window. Increase timeout to mitigate.

INFO

Missing OwnershipTransferred event in constructor

Standard off-chain indexing tools won't detect initial owner assignment.

// Auditor

Audit performed by eth-agents:auditor running on Claude Opus. Static analysis pass with manual pattern analysis + Slither scan.

// Token efficiency

Audit pipeline used ~14k tokens total. Prompt caching reduced cost by ~67% — static agent role definitions are cached after first invocation.

14k tokens used 67% cached
// Output

Every finding includes: severity classification, exact file location, description, impact statement, and a concrete remediation with code snippet. No vague suggestions.

// MCP HUD status
[audit] auditor ▶ Phase 2/5 · 14k tokens · 67% cached [audit] security-specialist ▶ Phase 3/5 · exploit sim idle — audit-report.md written
See the source on GitHub ↗

~55% token savings.
By design.

eth-agents is engineered to minimize token usage without sacrificing quality. Three complementary mechanisms work in layers.

~55% total savings
~35% model routing
~15% prompt caching
~12% output constraints
Layer Mechanism Savings
Model routing Haiku for simple tasks, Sonnet for implementation, Opus only for security and full-protocol pipelines ~35%
Prompt caching Static agent role definitions are cached after first invocation (each agent file exceeds Anthropic's 1024-token auto-cache threshold) ~15%
Output constraints Per-agent token limits, artifacts always before prose, structured outputs reduce verbose explanations ~12%

One command.
Zero config.

Install the plugin, navigate to any Ethereum project, and start describing what you want. eth-agents auto-detects Foundry or Hardhat, your frontend framework, and your Web3 library.

1
claude plugin install eth-agents
Install the plugin via Claude Code
2
cd your-eth-project
Navigate to any Foundry or Hardhat project
3
"audit this contract"
Describe what you want in plain English — that's it
View on GitHub Read the docs
Requires Claude Code · Node.js 18+ · MIT License