An AI gateway for the Blend Protocol on Stellar, enabling DeFi actions like lending, borrowing, and pool creation through AI assistants or apps.
Blend MCP is the universal AI gateway for the Blend Protocol on Stellar. It exposes all Blend DeFi actionsโlending, borrowing, pool creation, risk analysis, and moreโas simple, composable tools accessible by any AI assistant, bot, or app.
https://github.com/user-attachments/assets/008f02d5-1037-4b28-b1be-0af312f88de4
Blend MCP exposes the following tools (functions) to any AI assistant or app:
Tool Name | Description |
---|---|
loadPoolData | Loads comprehensive data for a given Blend pool (optionally includes user position). |
getTokenBalance | Gets the balance of a specific token for a user address. |
getFeeStats | Gets current Soroban network fee statistics. |
getPoolEvents | Gets historical events for a specific pool. |
loadBackstopData | Loads data for the main Blend backstop contract. |
loadTokenMetadata | Loads metadata for a given token/asset. |
simulateOperation | Simulates a transaction operation without submitting it. |
lend | Lend (supply collateral) to a pool. |
withdraw | Withdraw assets from a pool. |
borrow | Borrow assets from a pool. |
repay | Repay borrowed assets to a pool. |
claimRewards | Claim available rewards from a pool. |
createPool | Deploy a new, permissionless lending pool. |
addReserve | Add a new asset reserve to a pool. |
buyNft | Buy an NFT from a Soroban NFT contract using provided funds. |
Each tool is accessible via natural language or programmatic calls, and can be combined for advanced workflows.
Blend MCP's true power is in composabilityโAI assistants or apps can chain these tools to create advanced DeFi workflows:
loadPoolData
โ loadTokenMetadata
โ (optional) getPoolEvents
โ AI summarizes risk.borrow
โ buyNft
โ (optional) repay
later.loadPoolData
โ AI checks safety โ lend
if safe.createPool
โ addReserve
โ (optional) set parameters.flowchart TD
User(("User / AI Assistant"))
MCPServer(["Blend MCP Server"])
BlendSDK(["Blend Protocol SDK"])
Stellar(["Stellar Network"])
NFT(["NFT Contract / dApps"])
User-->|"Natural Language Command"|MCPServer
MCPServer-->|"Tool Call (lend, borrow, risk, etc.)"|BlendSDK
BlendSDK-->|"Soroban Transaction"|Stellar
MCPServer-->|"Composable Actions"|NFT
NFT-->|"On-chain Action"|Stellar
BlendSDK-->|"Data Fetch"|Stellar
MCPServer-->|"Result (Analysis, Confirmation, etc.)"|User
No install required! Use the published npm package with npx
:
{
"mcpServers": {
"Blend Protocol MCP": {
"command": "npx",
"args": [
"blend-mcp-stellar"
],
"env": {
"AGENT_SECRET": "<YOUR_STELLAR_SECRET_KEY>",
"RPC_URL": "https://soroban-testnet.stellar.org",
"BACKSTOP_ID_V2": "<OPTIONAL_BACKSTOP_ID_V2>",
"POOL_FACTORY_ID": "<OPTIONAL_POOL_FACTORY_ID>"
}
}
}
}
.cursor/mcp.json
(for most MCP clients) or claude_desktop_config.json
(for Claude Desktop).Run your AI assistant (Claude, ChatGPT, or any MCP-compatible client) and connect using the config above.
git clone [your-repo-url]
cd BlendMcp
yarn install # or npm install
AGENT_SECRET
(your Stellar secret key for signing transactions)POOL_FACTORY_ID
, BACKSTOP_ID
, etc. for advanced featuresnpm run build
# then run your MCP server as needed
claude_desktop_config.json
.Got 2 offers to sell this project ($2800 and $1900)
it is built for blend ecosystem
Blend MCP is under active development. More features and integrations are coming soon!
Built for the Stellar Blend hackathon, but designed for long-term extensibility and composability.
A tool to provide Large Language Models with project context by intelligently filtering and concatenating relevant files.
Access comprehensive documentation for Ant Design components, including examples, API references, and best practices.
Interact with the Moralis Web3 API to access blockchain data across multiple networks through a structured interface.
Provides AI-powered mentorship to LLM agents for tasks like code review, design critique, and brainstorming, using the Deepseek API.
A tool server for integrating Dify Workflows via the Model Context Protocol (MCP).
An MCP server for interacting with Juniper Junos network devices using LLMs.
An MCP (Model Context Protocol) aggregator that allows you to combine multiple MCP servers into a single endpoint allowing to filter specific tools.
Analyzes audio files and extracts metadata, tailored for game audio development workflows.
A Python client for connecting to Model Context Protocol (MCP) servers, supporting local scripts and npx packages.
A lightweight Java framework for building MCP servers with TCP transport via mcp-java-bridge.