Fluree MCP Server
A Node.js gateway that exposes FlureeDB functionality through MCP, enabling LLM clients to query, mutate, and manage Fluree ledgers interactively.
README
Fluree MCP Server
A modern Node.js gateway that exposes FlureeDB functionality through the Model Context Protocol (MCP). It combines a rich set of MCP tools, structured prompts, and session-aware routing so any LLM-powered client (IDE, chat, or custom app) can query, mutate, and manage a Fluree ledger safely and interactively.
✨ Key Features
| Category | Highlights |
|---|---|
| MCP Tools | 15+ tools: query, transact, block/block-range, history, network status, key generation, schema management (create / update / validate / export / import), database diagnostics, echo |
| Structured Prompts | Every tool ships with a matching prompt template; clients discover via prompts/list, retrieve with prompts/get, and enjoy argument completion support |
| Session Management | First initialize call establishes a session & stores connection info (dbUrl, network, ledger) so subsequent requests are stateless for the client |
| Validation & Safety | All tool inputs validated with Zod; errors surfaced as JSON-RPC -32602 or -32603 |
| Hot-Reload Friendly | Tools & prompts registered dynamically; add new modules under src/tools or src/prompts and restart |
| Extensible | Written in ES Modules, small surface area—drop in new handlers, middlewares, or MCP capabilities |
🗂️ Project Structure
src/
├─ config/ # Constant defaults & env helpers
├─ controllers/ # MCP server controller (session init, tool/prompt registration)
├─ routes/ # Express routes (`/api/v1/mcp`)
├─ tools/ # Individual MCP tools (each default export = { name, config, handler })
├─ prompts/ # Prompt templates mirrored one-to-one with tools
└─ index.js # Express bootstrap
.env.example # Sample environment variables
README.md # ← you are here
🚀 Quick Start
-
Clone & Install
git clone https://github.com/UDAIBHAT/Fluree_MCP_Server.git cd fluree-mcp-server npm install cp .env.example .env # adjust DB connection defaults if desired -
Run
npm start # serves http://localhost:3000 -
Initialize a Session
POST http://localhost:3000/api/v1/mcp Headers: { "Content-Type":"application/json" } Body: { "jsonrpc":"2.0", "id":1, "method":"initialize", "params":{ "dbUrl":"Your_DBurl", "network":"Your_Network", "ledger":"Your_Ledger" } }The response includes
mcp-session-id—add it to all subsequent requests. -
Discover Tools & Prompts
{ "jsonrpc":"2.0", "id":2, "method":"prompts/list", "params":{} }
🖥️ IDE & Non-IDE Client Setup
IDE clients
Cursor, Windsurf,Google Gemini . RooCode can add the server definition to their settings. Two examples are shown below – one for Cursor and one for Windsurf:
CURSOR IDE
- Just insert this Patch of Json-code under the MCP-Section in the Cursor settings and you are good to go.....
{
"mcpServers": {
"Fluree_MCP_Server": {
"url": "http://localhost:3000/api/v1/mcp",
"headers": {
"dburl": "Your_DBurl",
"network": "Your_Network",
"ledger": "Your_Ledger"
}
}
}
}
WINDSURF IDE
- Just insert this Patch of Json-code under the MCP-Section in the Windsurf Cascade settings and you are good to go.....
{
"mcpServers": {
"Fluree_MCP_Server": {
"serverUrl": "http://localhost:3000/api/v1/mcp",
"headers": {
"dbUrl": "Your_DBurl",
"network": "Your_Network",
"ledger": "Your_Ledger"
}
}
}
}
GOOGLE GEMINI
- Use the following JSON patch in your Gemini .settings
[ Currently Gemini doesnt support Streamable HTTP transport]
{
"mcpServers": {
"Fluree_MCP_Server": {
"httpUrl": "http://localhost:3000/api/v1/mcp?dburl=<Your_DBurl>&network=<Your_Network>&ledger=<Your_Ledger>",
"timeout": 30000,
"trust": true
}
}
}
Non-IDE clients
Can simply POST to /api/v1/mcp (or /mcp) with a JSON body containing dbUrl, network, and ledger as shown in the Quick-Start initialization request.
🔧 Tools Overview
| Name | Purpose | Input Parameters |
|---|---|---|
flureeQueryTool |
Select data via FlureeQL | selectFields (array|"*"), fromCollection |
flureeTransact |
Create random collections | count (1-50) |
flureeSchemaCreate |
Create collection + predicates | collectionName, predicates[], doc? |
flureeSchemaUpdate |
Update collection/predicates | collectionName, updateType, etc. |
| … | See src/tools for the rest |
Each tool handler returns { content:[ {type:"text"|"json", text:string|object} ] }, ready for LLM streaming.
💬 Prompt Workflow
- List →
method:"prompts/list" - Get →
method:"prompts/get", pass promptname+argumentsobject - (Optional) Suggest →
method:"completion/suggest"for auto-complete while filling arguments
Prompts use Mustache-style placeholders ({{argument}}) inside user messages to instruct the assistant to call the correct tool.
⚙️ Environment Variables
| Variable | Default | Description |
|---|---|---|
PORT |
3000 |
Server port |
FLUREE_DB_URL |
http://localhost:8090 |
Base Fluree URL |
FLUREE_NETWORK |
test |
Default network |
FLUREE_LEDGER |
demo |
Default ledger |
📑 API Collection
A ready-to-import Postman collection lives at docs/postman/Fluree_MCP.postman_collection.json (or generate it from the sample snippet in this README).
-
You can also make a new collection with specified MCP endpoint ,postman supports the mcp endpoints so just inseret the server url then it will give you the whole list of tools.
-
Just Import the collection and call the
Initialize_reqthis will make Initialization_Connection to the MCP_Server and you can also change theparamslikedbUrl,network,ledger -
The mcp-session-id is dynamically stored in every request through Postman Post-Req Scripts
🛣️ Roadmap / Nice-to-Haves
- JWT authentication & role-based prompt permissions
- Rate limiting & CORS controls
- WebSocket support for real-time block watch
- Production observability: Prometheus exporter & Grafana dashboard
- Docker Compose bundle (Fluree + MCP + Grafana)
📜 License
MIT © UDAI BHAT
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。