datacloud-code-mcp

datacloud-code-mcp

A Salesforce Data Cloud MCP server using Cloudflare Code Mode, enabling agents to discover and execute Data Cloud operations via a two-step search-and-execute loop with JavaScript code.

Category
访问服务器

README

Data Cloud Code MCP

A Salesforce Data Cloud (Data360) MCP server using the Cloudflare Code Mode pattern: two tools (search + execute), each taking a JavaScript async arrow function, with a fixed ~1k-token tool footprint regardless of API surface size.

How It Works

Instead of exposing hundreds of endpoint-level tools, this server gives agents a stable two-step loop:

  1. search — agent writes JavaScript to filter the OpenAPI spec and discover Data Cloud endpoints.
  2. execute — agent writes JavaScript that calls salesforce.request() to make authenticated API requests.

Both tools run user code in a node:vm sandbox with restricted globals. The full OpenAPI spec never enters the model context — the agent explores it programmatically through search().

What's Covered

~185 Data Cloud operations across 25 API families:

Family Examples
Query SQL query (v1/v2/v3), profile, insights, data graphs
DMO/DLO CRUD for Data Model Objects and Data Lake Objects
Mappings DMO field mappings, bulk mappings, standard mappings
Data Streams List, create, update, delete, trigger runs
Connectors List types, metadata, CRUD connections, test
Calculated Insights CRUD, run, validate, enable/disable
Segments CRUD, publish, members, overlap analysis
Identity Resolution CRUD rulesets, publish, run, match history
Activations CRUD activations + targets, sync, history
Data Transforms CRUD, run, validate, schedule
Semantic Data Models Models, objects, dimensions, measurements, metrics, relationships, query
Data Spaces CRUD spaces, member management
DataKits List, manifest, deploy, undeploy, component status
GDPR Right-to-access, bulk read, erasure requests
Search Indexes CRUD, hybrid full-text query
Eventing Single and batch a360 event publish
Data Actions CRUD actions + targets

Quick Start

npm install
npm run build
npm test        # 27 tests across 5 suites

Auth Options

Option 1: Direct access token (same env vars as d360-mcp-server)

CDP_ACCESS_TOKEN=<token> CDP_INSTANCE_URL=<url> npm run dev

Option 2: OAuth client credentials

CDP_CLIENT_ID=<id> CDP_CLIENT_SECRET=<secret> CDP_LOGIN_URL=https://login.salesforce.com npm run dev

Option 3: OAuth web flow

  1. Set SALESFORCE_OAUTH_CLIENT_ID, SALESFORCE_OAUTH_CLIENT_SECRET, SALESFORCE_OAUTH_REDIRECT_URI in .env.
  2. Start server: npm run dev
  3. Open /oauth/start?user_id=default and complete login.

Option 4: Seed token from CLI

npm run seed:token     # reads sf CLI auth or env vars
TOKEN_STORE_PATH=./data/tokens.integration.json \
TOKEN_ENCRYPTION_KEY_BASE64='<from seed>' npm run dev

Run Server

PORT=3000 HOST=127.0.0.1 npm run dev
curl -sS http://127.0.0.1:3000/healthz

Smoke Test

MCP_URL=http://127.0.0.1:3000/mcp USER_ID=default npm run smoke:mcp

Example Calls

Search: find endpoints by tag

{
  "name": "search",
  "arguments": {
    "code": "async () => {\n  const results = [];\n  for (const [path, methods] of Object.entries(spec.paths)) {\n    for (const [method, op] of Object.entries(methods)) {\n      if (op.tags?.some(t => t.toLowerCase().includes('segment'))) {\n        results.push({ method: method.toUpperCase(), path, summary: op.summary });\n      }\n    }\n  }\n  return results;\n}"
  }
}

Search: inspect an endpoint schema

{
  "name": "search",
  "arguments": {
    "code": "async () => {\n  const op = spec.paths['/services/data/v64.0/ssot/query-sql']?.post;\n  return { summary: op?.summary, requestBody: op?.requestBody };\n}"
  }
}

Execute: run a SQL query

{
  "name": "execute",
  "arguments": {
    "code": "async () => {\n  return await salesforce.request({\n    method: 'POST',\n    path: '/services/data/v64.0/ssot/query-sql',\n    body: { sql: 'SELECT FirstName__c FROM UnifiedIndividual__dlm LIMIT 5' }\n  });\n}"
  }
}

Execute: chain multiple calls

{
  "name": "execute",
  "arguments": {
    "code": "async () => {\n  const list = await salesforce.request({ method: 'GET', path: '/services/data/v64.0/ssot/segments' });\n  const first = list.body?.data?.[0];\n  if (!first) return { message: 'No segments' };\n  return await salesforce.request({ method: 'GET', path: '/services/data/v64.0/ssot/segments/' + first.id });\n}"
  }
}

MCP Client Integration

{
  "mcpServers": {
    "datacloud-code-mcp": {
      "transport": "streamable_http",
      "url": "http://127.0.0.1:3000/mcp",
      "headers": { "x-user-id": "default" }
    }
  }
}

Safety Model

  • Mutating methods (POST, PATCH, PUT, DELETE) blocked unless ALLOW_WRITES=true.
  • salesforce.request() only allows outbound HTTP to the authenticated instance's hostname + *.salesforce.com + *.force.com.
  • Sensitive headers/body keys are redacted in tool output.
  • User code runs in a node:vm sandbox with no require, process, global, or filesystem access.
  • Sandbox enforces SANDBOX_TIMEOUT_MS (default 15s) to prevent runaway execution.

Deploy to Heroku

heroku create
heroku config:set TOKEN_ENCRYPTION_KEY_BASE64=$(node -e "console.log(require('crypto').randomBytes(32).toString('base64'))")
heroku config:set ALLOW_WRITES=false
heroku config:set CDP_ACCESS_TOKEN=<token> CDP_INSTANCE_URL=<url>
git push heroku main

Then point your MCP client at https://<app>.herokuapp.com/mcp.

Configuration

See .env.example for all available environment variables.

Project Layout

src/
  index.ts              HTTP server + MCP session management
  mcp-server.ts         Tool registration (search, execute, auth_status)
  config.ts             Zod-parsed environment config
  types.ts              Shared TypeScript interfaces
  logger.ts             Pino logger
  auth/
    oauth-service.ts    Salesforce OAuth web flow + token exchange
    auth-modes.ts       Strategy resolver (direct/client-cred/password/oauth)
    token-store.ts      AES-256-GCM encrypted token persistence
  schema/
    datacloud-schema-service.ts   Catalog loader, merger, $ref resolver
    catalog.ts                    OpenAPI → PlatformOperation parser
    spec-processor.ts             $ref resolution + spec processing (Cloudflare pattern)
    bundled-spec.ts               Resolve bundled YAML paths
    data360-api.bundled.yaml      Base OpenAPI spec (~35 endpoints)
    d360-extras.yaml              Extended endpoints (~150 more operations)
  sandbox/
    runner.ts           node:vm sandbox executor
    sf-client.ts        salesforce.request() injectable client
    safe-fetch.ts       Hostname-allow-listed fetch wrapper
    truncate.ts         Response truncation
  execute/
    datacloud-executor.ts   Legacy structured executor (kept for reference)
    redaction.ts            Body/header redaction helpers
    http-policy.ts          Retry + read-cache helpers
  safety/
    write-confirmation.ts   HMAC write tokens (used by legacy executor)
  search/
    search-index.ts         BM25 search index (kept for potential reuse)
  utils/
    crypto.ts               AES-256-GCM encrypt/decrypt
    headers.ts              Header value resolver
tests/                      Vitest test suites
scripts/                    Smoke test + token seed scripts
docs/                       Reference docs (Data Cloud guide, Postman)

References

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选