Zoho FSM MCP Server
Enables AI assistants to interact with Zoho Field Service Management, allowing operations on work orders, requests, contacts, and more via natural language.
README
Zoho FSM MCP Server
A production-ready Model Context Protocol server that exposes the Zoho FSM (Field Service Management) REST API as AI-friendly tools, resources, and prompts.
It is modular and extensible: thin API-wrapper tools, higher-level "intelligent" workflow tools, cached read-only resources, and a dedicated extension point for tools auto-generated from an OpenAPI spec.
Features
- 🔐 OAuth with automatic access-token refresh (refresh-token grant), multi-region.
- 🌐 Single API client (
FSMClient) with retries, backoff, timeouts, 401 recovery, and typed responses. - 🧰 Tools for Requests, Work Orders, Appointments, Contacts, Companies, Estimates, Invoices, Assets, and Users.
- 🤖 Intelligent workflow tools that orchestrate multiple API calls (
create_service_request,assign_best_technician,complete_job). - 📚 Resources (
fsm://modules,fsm://statuses,fsm://territories,fsm://users,fsm://services,fsm://parts,fsm://metadata) with TTL caching. - 💬 Prompt templates (
dispatch-summary,job-summary,invoice-summary,technician-brief,customer-history). - 🪵 Structured logging to stderr (never logs secrets) and centralized error handling.
- 🧩 Extensible:
src/tools/generated/is reserved for OpenAPI-generated tools.
Project structure
zoho-fsm-mcp/
├── src/
│ ├── auth/oauth.ts # OAuth token manager (refresh + caching)
│ ├── client/
│ │ ├── fsmClient.ts # The only place HTTP happens
│ │ └── types.ts # Shared FSM/response types
│ ├── tools/
│ │ ├── requests.ts workOrders.ts appointments.ts contacts.ts
│ │ ├── companies.ts invoices.ts estimates.ts assets.ts users.ts
│ │ ├── intelligent.ts # Multi-step workflow tools
│ │ ├── generated/ # Reserved for OpenAPI-generated tools
│ │ ├── shared.ts # Shared Zod shapes + context type
│ │ └── index.ts # registerAllTools()
│ ├── resources/ # modules, statuses, metadata (+ live)
│ ├── prompts/ # Reusable prompt templates
│ ├── utils/ # config, logger, errors, cache, mcp helpers
│ ├── server.ts # Wires everything together
│ └── index.ts # stdio entry point
├── .env.example
├── package.json tsconfig.json eslint.config.js .prettierrc
└── README.md
Installation
git clone <this-repo>
cd zoho-fsm-mcp
npm install
npm run build
Requires Node.js ≥ 18.
OAuth setup
- Go to the Zoho API Console and create a Self Client (or Server-based Application).
- Note the Client ID and Client Secret.
- Generate a grant token with the FSM scopes, e.g.:
ZohoFSM.modules.ALL,ZohoFSM.settings.ALL,ZohoFSM.users.READ - Exchange the grant token for a refresh token (one-time), using the accounts endpoint for your region:
Save thecurl -X POST "https://accounts.zoho.com/oauth/v2/token" \ -d "grant_type=authorization_code" \ -d "client_id=YOUR_CLIENT_ID" \ -d "client_secret=YOUR_CLIENT_SECRET" \ -d "code=YOUR_GRANT_TOKEN"refresh_tokenfrom the response. - Copy
.env.exampleto.envand fill in:ZOHO_CLIENT_ID=... ZOHO_CLIENT_SECRET=... ZOHO_REFRESH_TOKEN=... ZOHO_REGION=com # com | eu | in | au | jp
The server refreshes access tokens automatically and picks the correct base URL from ZOHO_REGION:
| Region | Accounts endpoint | FSM API base |
|---|---|---|
com |
accounts.zoho.com |
fsm.zoho.com/fsm/v1 |
eu |
accounts.zoho.eu |
fsm.zoho.eu/fsm/v1 |
in |
accounts.zoho.in |
fsm.zoho.in/fsm/v1 |
au |
accounts.zoho.com.au |
fsm.zoho.com.au/fsm/v1 |
jp |
accounts.zoho.jp |
fsm.zoho.jp/fsm/v1 |
Running locally
# Development (auto-reload)
npm run dev
# Type-check / lint / format
npm run typecheck
npm run lint
npm run format
# Production
npm run build
npm start
Inspect the tools interactively with the MCP Inspector:
npm run inspect
Configuring Claude Desktop
Edit claude_desktop_config.json:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"zoho-fsm": {
"command": "node",
"args": ["/absolute/path/to/zoho-fsm-mcp/dist/index.js"],
"env": {
"ZOHO_CLIENT_ID": "...",
"ZOHO_CLIENT_SECRET": "...",
"ZOHO_REFRESH_TOKEN": "...",
"ZOHO_REGION": "com"
}
}
}
}
Restart Claude Desktop. The zoho-fsm tools appear in the tools menu.
Configuring ChatGPT / other MCP clients
Any MCP-compatible client that supports stdio servers can launch it the same way:
node /absolute/path/to/zoho-fsm-mcp/dist/index.js
For ChatGPT's MCP support, register a connector pointing at this command (or wrap it behind an HTTP/SSE bridge if your client requires a URL). Environment variables are supplied the same way as above.
Available tools
| Category | Tools |
|---|---|
| Requests | create_request, get_request, search_requests, update_request |
| Work Orders | create_work_order, update_work_order, search_work_orders |
| Appointments | create_appointment, update_appointment, schedule_appointment |
| Contacts | create_contact, search_contacts |
| Companies | create_company, search_companies |
| Estimates | create_estimate |
| Invoices | create_invoice, mark_invoice_paid |
| Assets | create_asset, update_asset |
| Users | list_users, get_user |
| Workflows | create_service_request, assign_best_technician, complete_job |
Every tool validates input with Zod, calls FSMClient, and returns a structured MCP response. Errors are converted into a consistent payload with status, code, retryable, message, and details.
Adding a new tool
- Create (or extend) a file in
src/tools/, e.g.parts.ts. - Export a
registerXTools(ctx: ServerContext)function. - Inside it, call
server.registerTool(name, { title, description, inputSchema }, handler).- Define
inputSchemaas a Zod raw shape. - Wrap the handler with
withToolLogging(name, ...)for logging + error handling. - Build the API payload with
buildRecord(...)and call aFSMClientmethod. - Return via
ok(data, summary).
- Define
- Register your function in
src/tools/index.ts.
Example skeleton:
export function registerPartTools({ server, client }: ServerContext): void {
server.registerTool(
'create_part',
{ title: 'Create Part', description: '...', inputSchema: { name: z.string() } },
withToolLogging('create_part', async (args) => {
const created = await client.create('Parts', buildRecord({ Name: args.name }));
return ok(created, `Created part "${args.name}".`);
}),
);
}
New REST calls should go through FSMClient (add a method there) — never call axios directly from a tool.
Future: OpenAPI-generated tools
src/tools/generated/ is reserved for tools generated from the Zoho FSM OpenAPI spec. A codegen step will emit *.generated.ts files there, each exporting a register…GeneratedTools(ctx) function called from generated/index.ts. Generated code stays separate from the hand-written intelligent tools and is safe to regenerate wholesale.
License
MIT
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。