Soroban MCP Server
Enables AI agents to build, test, deploy, invoke, and inspect Soroban smart contracts on the Stellar network through structured MCP tools and resources.
README
Soroban MCP Server
An MCP (Model Context Protocol) server that gives AI coding agents direct, structured access to Soroban smart contracts and the Stellar network — so agents can build, test, deploy, and inspect contracts without shelling out blindly to the CLI and parsing raw text output.
Demo: (drop a terminal recording or gif here once you have a working build)
Why this exists
AI coding agents are increasingly used to write and debug smart contracts, but today they either guess at CLI syntax or parse unstructured terminal output to decide whether a build succeeded, what a contract's state looks like, or why a deployment failed.
This server exposes those operations as typed, validated MCP tools with structured results and clear error messages — making agent-assisted Soroban development faster and more reliable.
Features
| Tool | Description |
|---|---|
build_contract |
Builds a Soroban project; returns structured compiler diagnostics (file, line, error code, message) |
run_tests |
Runs cargo test; returns a structured pass/fail report with failure messages |
deploy_contract |
Deploys a compiled .wasm to testnet or futurenet |
invoke_contract |
Simulates a contract function call; returns result + resource cost (CPU, memory, fee) |
get_contract_state |
Reads a contract's persistent/instance/temporary storage entries |
get_account_info |
Returns balance, sequence number, and signers for a Stellar account |
Resources exposed:
soroban://network/{network}/status— current ledger height and healthsoroban://contract/{id}/abi— contract ABI/spec from on-chain WASM metadatasoroban://contract/{id}/transactions— recent transaction history
Installation
npm install -g soroban-mcp-server
Or run directly without installing:
npx soroban-mcp-server
Prerequisites:
- Node.js ≥ 20
- stellar-cli on your
PATH(required forbuild_contractanddeploy_contract)
Configuration
Add to your MCP client config (e.g. Claude Code's mcp.json):
{
"mcpServers": {
"soroban": {
"command": "npx",
"args": ["soroban-mcp-server"],
"env": {
"STELLAR_NETWORK": "testnet"
}
}
}
}
| Env var | Description | Default |
|---|---|---|
STELLAR_NETWORK |
testnet, futurenet, or mainnet (read-only ops only on mainnet) |
testnet |
RPC_URL |
Override the default RPC endpoint | network default |
SOROBAN_CLI_PATH |
Path to the Soroban/Stellar CLI binary | stellar |
LOG_LEVEL |
debug, info, warn, error |
info |
Usage example
Once connected to your MCP client, an agent can run a full build-test-deploy cycle:
"Build this contract, run the tests, and if they pass, deploy it to testnet and give me the contract ID."
The agent calls build_contract → run_tests → deploy_contract in sequence, using the structured output of each step to decide whether to proceed — without you needing to copy-paste terminal output back and forth.
Tool call examples
Build a contract:
{
"tool": "build_contract",
"arguments": { "project_path": "./contracts/my_token" }
}
Run tests with a filter:
{
"tool": "run_tests",
"arguments": { "project_path": "./contracts/my_token", "filter": "test_transfer" }
}
Check an account:
{
"tool": "get_account_info",
"arguments": { "public_key": "GABC...", "network": "testnet" }
}
Invoke a contract function:
{
"tool": "invoke_contract",
"arguments": {
"contract_id": "CXXX...",
"function_name": "balance",
"args": [{ "type": "address", "value": "GABC..." }],
"network": "testnet"
}
}
Error handling
Every tool returns structured errors with a code field instead of raw stack traces:
{
"error": "Insufficient balance to complete the transaction. Fund the account at https://friendbot.stellar.org",
"code": "INSUFFICIENT_BALANCE"
}
Known error codes: BUILD_FAILED, WASM_NOT_FOUND, INVALID_WASM, NETWORK_TIMEOUT, NETWORK_UNREACHABLE, RPC_ERROR, INVALID_CONTRACT_ID, CONTRACT_NOT_FOUND, CONTRACT_INVOCATION_FAILED, INSUFFICIENT_BALANCE, ACCOUNT_NOT_FOUND, INVALID_PUBLIC_KEY, CLI_NOT_FOUND, CLI_EXECUTION_FAILED, TEST_EXECUTION_FAILED.
Development
git clone https://github.com/your-org/soroban-mcp-server
cd soroban-mcp-server
npm install
npm run typecheck # TypeScript strict check
npm run lint # ESLint
npm test # Unit tests (all network/CLI calls mocked)
# Live testnet integration tests (requires network access)
INTEGRATION=true npm run test:integration
Roadmap
This project is built to grow alongside the Soroban ecosystem. Planned areas of expansion (and a source of ongoing, scoped contribution issues):
- TTL management —
bump_contract_ttltool to extend contract state lifetimes - Signing support — submit state-changing transactions, not just simulate them
- Additional wallets/signers — Freighter, hardware wallets
- Richer error diagnosis — new failure modes as the protocol evolves
- Prompt templates — common contract patterns (token, escrow, multisig)
- Local devnet support —
quickstartDocker integration for offline development
Contributing
See CONTRIBUTING.md. Issues are labeled by scope and complexity — issues tagged good-first-issue are a great place to start.
License
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。