@zhoujinandrew/mcp-te-server-v2

MCP (Model Context Protocol) server for ThinkingEngine (TE) analytics platform. Enables AI assistants like Claude to query events, run SQL, manage dashboards, user segments, and operations directly through TE's API.

Quick Setup

Add to your MCP client configuration (e.g. Claude Code ~/.claude.json, Cursor, etc.):

{
  "mcpServers": {
    "mcp-te-v2": {
      "command": "npx",
      "args": ["-y", "@zhoujinandrew/mcp-te-server-v2@latest"],
      "env": {
        "TE_HOST": "ta.thinkingdata.cn"
      }
    }
  }
}

Prerequisites

• Node.js >= 18 (for native fetch support)
• A TE platform account with access to the target project

macOS Auto-Auth (Optional, Recommended)

For zero-interaction token capture from Chrome:

1. Open Chrome → View → Developer → Allow JavaScript from Apple Events (one-time toggle)
2. macOS will prompt for Automation permission on first run — click Allow

Authentication Flow

The server uses a 4-step auth cascade. No manual configuration needed in most cases:

| Step | What happens | User action | |------|-------------|-------------| | 1. Cache | Reads ~/.te-mcp/token.json | None | | 2. osascript (macOS) | Extracts ACCESS_TOKEN from an open TE Chrome tab | None | | 3. Open & Poll (macOS) | Opens TE login page, polls Chrome until token appears | Login to TE | | 4. Manual Fallback | Opens a local page with a text input for pasting token | Paste token from DevTools |

Token is cached at ~/.te-mcp/token.json and reused until it expires (401/403 triggers re-auth automatically).

Manual Token Reset

rm ~/.te-mcp/token.json

Environment Variables

| Variable | Required | Default | Description | |----------|----------|---------|-------------| | TE_HOST | Yes | ta.thinkingdata.cn | TE platform hostname |

Available Tools

Metadata

| Tool | Description | |------|-------------| | te_list_events | Get event catalog for a project | | te_load_event_props | Load filterable properties for events | | te_load_measure_props | Load measurable/aggregatable properties | | te_list_entities | List analysis entities (user, account, device, etc.) | | te_list_metrics | List predefined metrics | | te_list_tables | List SQL-queryable tables | | te_get_table_columns | Get column definitions for a SQL table |

Analysis

| Tool | Description | |------|-------------| | te_list_reports | List all reports | | te_get_report | Get report definition details | | te_save_report | Create or update a report | | te_list_dashboards | List all dashboards | | te_get_dashboard | Get dashboard details | | te_create_dashboard | Create a new dashboard | | te_update_dashboard | Update dashboard layout | | te_list_dashboard_reports | List reports in a dashboard | | te_query_report_data | Query report data via WebSocket | | te_query_sql | Execute SQL queries |

Audience & Segments

| Tool | Description | |------|-------------| | te_list_tags | List user tags | | te_get_tag | Get tag details and rules | | te_list_clusters | List user clusters/segments | | te_predict_cluster_count | Predict entity count for cluster conditions | | te_list_audience_events | List events for audience targeting | | te_load_audience_props | Load audience targeting properties |

Operations

| Tool | Description | |------|-------------| | te_create_task | Create and submit an operation task | | te_list_tasks | List operation tasks | | te_get_task_stats | Get task statistics | | te_save_flow | Create or update a flow canvas | | te_list_flows | List flow canvases | | te_get_flow | Get flow definition | | te_list_channels | List push channels | | te_get_channel | Get channel details |

Helpers

| Tool | Description | |------|-------------| | te_get_space_tree | Get navigation space tree | | te_get_timezone | Get project timezone offset | | te_list_mark_times | List date annotations/milestones |

Development

git clone https://github.com/zjandrew/mcp-te-server-v2.git
cd mcp-te-server-v2
npm install
TE_HOST=ta.thinkingdata.cn node src/index.js

License

MIT