foundry-zksync-mcp
Exposes foundry-zksync CLI tools to AI assistants, enabling smart contract development on zkSync with key management options.
README
foundry-zksync-mcp
An MCP (Model Context Protocol) server that exposes foundry-zksync CLI tools to AI assistants like Claude.
Tools
| Tool | Description |
|---|---|
init |
forge init with automatic zkSync config |
compile |
forge build --zksync |
test |
forge test --zksync with filter/verbosity options |
run_script |
forge script --zksync with broadcast, sender, slow mode |
deploy |
forge create --zksync with structured output parsing |
verify |
forge verify-contract --zksync (Etherscan & zkSync Explorer) |
install |
forge install for dependency management |
clean |
forge clean to remove build artifacts |
gas_report |
forge test --gas-report for gas usage analysis |
snapshot |
forge snapshot for gas benchmarking |
read_foundry_toml |
Read and inspect project configuration |
explain |
Match errors/logs against a knowledge base of zkSync gotchas |
cast_abi_encode |
cast abi-encode |
cast_abi_decode |
cast abi-decode |
cast_calldata_decode |
cast calldata-decode |
cast_call |
cast call for read-only contract interaction |
cast_send |
cast send for state-changing transactions |
cast_balance |
cast balance for ETH balance queries |
cast_nonce |
cast nonce for transaction nonce queries |
get_zksync_docs |
Look up foundry-zksync documentation by topic |
anvil_zksync |
Start or check a local anvil-zksync dev node |
Prerequisites
- Node.js >= 18
- foundry-zksync installed (
foundryup-zksync)
Quick Start (npx)
No install needed — run directly with npx:
claude mcp add foundry-zksync -- npx -y foundry-zksync-mcp
Or add to claude_desktop_config.json / ~/.claude.json:
{
"mcpServers": {
"foundry-zksync": {
"command": "npx",
"args": ["-y", "foundry-zksync-mcp"]
}
}
}
Install from Source
git clone https://github.com/Jrigada/foundry-zksync-mcp.git
cd foundry-zksync-mcp
npm install
npm run build
Then register with Claude Code:
claude mcp add foundry-zksync node /absolute/path/to/foundry-zksync-mcp/dist/index.js
Key Management
All signing tools (deploy, cast_send, run_script) support multiple wallet methods. Choose based on your security needs:
Local Development (anvil-zksync)
For local dev with well-known test keys, use privateKey directly:
privateKey: "0xac0974bec39a17e36ba4a6b4d238ff944bacb478cbed5efcae784d7bf4f2ff80"
These are the default anvil-zksync test accounts — they hold no real value.
Warning: Never use
privateKeywith keys that hold real funds. MCP tool parameters are visible to the AI assistant and may be logged.
Recommended for Production: Named Keystores
Keys are encrypted on disk — only the account name travels through MCP.
# Import a private key into a named keystore (interactive, key never shown)
cast wallet import deployer --interactive
# List your keystores
ls ~/.foundry/keystores/
Then use account: "deployer" in any signing tool. Forge will prompt for the password at runtime, or you can point to a password file with passwordFile.
Keystore Files
If you have an existing encrypted keystore JSON file (e.g. from Geth, MetaMask export):
keystore: "/path/to/keystore.json"
passwordFile: "/path/to/password.txt"
Hardware Wallets
ledger: true # Ledger
trezor: true # Trezor
Cloud KMS
aws: true # AWS KMS (set AWS_KMS_KEY_ID env var)
gcp: true # Google Cloud KMS (set GCP_PROJECT_ID, GCP_LOCATION, etc.)
Signing Method Summary
| Method | Key Exposure | Best For |
|---|---|---|
| Hardware wallet | None (key never leaves device) | High-value production |
| Cloud KMS | None (key in HSM) | Automated production |
Named keystore (account) |
None through MCP (encrypted on disk) | General production use |
| Keystore file | None through MCP (encrypted on disk) | Existing workflows |
privateKey |
Visible to AI assistant | Local dev with test keys only |
Project Structure
src/
index.ts MCP server entry point, registers all 21 tools
knowledge.ts Knowledge base (45+ entries) and system addresses
tools/
shared.ts Profile field and env builder shared across forge tools
init.ts forge init with zkSync config
compile.ts forge build --zksync
test.ts forge test --zksync
run_script.ts forge script --zksync
deploy.ts forge create --zksync (structured output)
verify.ts forge verify-contract --zksync
install.ts forge install
clean.ts forge clean
gas_report.ts forge test --gas-report
snapshot.ts forge snapshot
read_foundry_toml.ts Read project config
explain.ts Error/log explanation with knowledge base
get_zksync_docs.ts Documentation lookup with embedded content
cast_abi_encode.ts cast abi-encode
cast_abi_decode.ts cast abi-decode
cast_calldata_decode.ts cast calldata-decode
cast_call.ts cast call
cast_send.ts cast send
cast_balance.ts cast balance
cast_nonce.ts cast nonce
anvil_zksync.ts anvil-zksync node management
__tests__/ 133 tests (vitest)
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。