Dakota Marketplace MCP Server
Enables AI clients to query Dakota Marketplace data by integrating directly with Salesforce via OAuth and PKCE. Users can search and retrieve detailed profiles for fund managers, allocators, and funds while maintaining native Salesforce security and sharing rules.
README
Dakota Marketplace MCP Server
External-facing MCP (Model Context Protocol) server that allows customers to query Dakota Marketplace data from any MCP-compatible AI client. Every query runs as the authenticated customer using their Salesforce OAuth token — Salesforce sharing rules and field-level security are enforced natively.
Architecture
AI Client (Claude/ChatGPT/Cursor)
↓ MCP Protocol
Dakota MCP Server (Node.js + Express)
├─ OAuth 2.0 + PKCE → Salesforce Connected App
├─ Redis → Session storage, cache, rate limiting
└─ Salesforce REST/Composite API → Per-user queries
MCP Tools
| Tool | Description |
|---|---|
search_fund_managers |
Search by name, AUM, strategy, geography |
get_fund_manager |
Full profile + funds + contacts (Composite API) |
search_allocators |
Search by type, AUM, location |
get_allocator |
Full allocator profile + contacts |
search_funds |
Search by strategy, vintage year, status, manager |
get_fund |
Full fund details + manager info |
get_contacts |
Contacts for a manager or allocator |
search_marketplace |
Free-text search across all entities (SOSL) |
Setup
1. Salesforce Connected App
- In Salesforce Setup → App Manager → New Connected App
- Enable OAuth Settings:
- Callback URL:
https://your-app.herokuapp.com/auth/callback - OAuth Scopes:
api,refresh_token
- Callback URL:
- Enable PKCE: check "Require Proof Key for Code Exchange (PKCE)"
- Set "Permitted Users" to "All users may self-authorize"
- Note the Consumer Key and Consumer Secret
2. Environment Variables
Copy .env.example to .env and fill in:
cp .env.example .env
| Variable | Description |
|---|---|
SALESFORCE_CLIENT_ID |
Connected App Consumer Key |
SALESFORCE_CLIENT_SECRET |
Connected App Consumer Secret |
SALESFORCE_INSTANCE_URL |
e.g., https://dakota.my.salesforce.com |
SALESFORCE_REDIRECT_URI |
e.g., https://your-app.herokuapp.com/auth/callback |
REDIS_URL |
Redis connection string |
SESSION_SECRET |
Random secret for session signing |
PORT |
Server port (Heroku sets automatically) |
3. Install & Run
npm install
npm start
4. Deploy to Heroku
heroku create dakota-mcp-server
heroku addons:create heroku-redis:mini
heroku config:set SALESFORCE_CLIENT_ID=... SALESFORCE_CLIENT_SECRET=... ...
git push heroku main
Authentication Flow
- Customer opens
GET /auth/login— redirects to Salesforce OAuth - After login, Salesforce redirects to
/auth/callbackwith authorization code - Server exchanges code for access + refresh tokens (PKCE verified)
- Tokens stored in Redis keyed by session ID
- Session ID returned to client for use in MCP tool calls
- Token refresh happens automatically before expiry
Rate Limits
- Standard tier: 100 requests/hour
- Premium tier: 500 requests/hour
- Sliding window tracked in Redis
- Clear error with retry guidance when exceeded
Cache Strategy
- Search results: 120s TTL
- Profile lookups: 300s TTL
- Strictly isolated by user ID — no cross-customer data leakage
- Key format:
cache:{user_id}:{tool_name}:{query_hash}
Customizing SOQL Queries
All Salesforce field names are centralized in the tool files under src/tools/. Update the SOQL queries there to match your actual Salesforce object and field API names. The Composite API request builders are in src/salesforce/composite.js.
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。