Skip to content
MCP Server Directory logo
developer-tools

Agenticbi Mcp Server

Connect Claude, ChatGPT, Gemini, and other MCP-enabled AI clients directly to your business data using AgenticBI.

Reviewed overview

MCP 服务器概览

What is AgenticBI MCP Server? AgenticBI MCP Server lets MCP-enabled AI clients (Claude, ChatGPT, Gemini-compatible connectors, Cursor, VS Code, Codex CLI and others) connect directly to your business data and workflows. It implements the MCP 2024-11-05 protocol over streamable HTTP (POST) with SSE and exposes 33 AI-driven tools for querying, managing, exporting, ingesting, and acting on data.

How to use the project?

  • Endpoint: https://app.agenticbi.com/api/2.0/mcp
  • Authentication: OAuth 2.1 with PKCE (or manual MCP token via AgenticBI → Settings → AI Settings → MCP Token).
  • Client setup is done by adding the AgenticBI MCP server URL (or bearer token) to your client configuration:
    • Claude Desktop: add to claude_desktop_config.json
    • Claude Code: run claude mcp add agenticbi --transport http https://app.agenticbi.com/api/2.0/mcp
    • Cursor: add to mcp.json via Tools & Integrations
    • VS Code: use Command Palette → MCP: Add Server or add to settings.json/.vscode/mcp.json
    • ChatGPT: Add Custom Connector (OAuth 2.0 or Bearer Token)
    • Codex CLI: add to ~/.codex/config.json
  • On first use clients open a browser for AgenticBI sign-in/authorization. Full guides available at https://app.agenticbi.com/docs/mcp-server.html and per-client docs linked in the project.

Key features

  • Direct, secure connection from MCP clients to business data using MCP 2024-11-05
  • Streamable HTTP with Server-Sent Events for realtime responses
  • OAuth 2.1 with PKCE and optional manual token authentication
  • 33 tools across categories: Ask & Query, Dashboards & Exports, Insights & Alerts, Documents, Create & Manage, Destructive
  • Tools include: natural-language ask, get_data, search, run saved queries, create/update dashboards/widgets/reports/alerts, push_data, import_file, ingest/extract document data, export_pdf/export_csv, delete ops
  • Cloud9QL assist for query generation/validation
  • Health & limits: idle timeout 30 min, max concurrent sessions 1,000, default/max row limits (defaults: ask 500, get_data 10,000; max 200,000)

Use cases

  1. Ask business questions in natural language (e.g., top 10 customers by revenue last quarter).
  2. Build, update and layout dashboards and visualizations programmatically.
  3. Export dashboards or data as PDF/CSV and distribute to teams.
  4. Create alerts and test notifications (e.g., MRR drop alerts).
  5. Ingest PDFs/transcripts and extract structured metrics into datasets.
  6. Run saved queries and scheduled reports, push rows into datasets, or create data sources.
  7. Generate and validate Cloud9QL queries and get AI-powered insights.

FAQ (from the project)

  • Which clients are supported?

MCP-compatible clients such as Claude Desktop/Code, Cursor, VS Code, ChatGPT connectors, and Codex CLI are supported.

  • How do I authenticate?

Use OAuth 2.1 with PKCE for standard flows. Alternatively generate an MCP token in AgenticBI settings and supply it as a Bearer token in client configuration.

  • What is the server endpoint and protocol?

Endpoint: https://app.agenticbi.com/api/2.0/mcp. Protocol: MCP 2024-11-05 over streamable HTTP (POST) with SSE.

  • How many tools and what capabilities are available?

33 tools across querying, dashboards/exports, insights/alerts, document QA, creation/management, and destructive operations.

  • What are important limits/health details?

Idle timeout: 30 minutes. Max concurrent sessions: 1,000. Default rows: ask 500, get_data 10,000. Max rows: 200,000. Health endpoint: GET /api/2.0/mcp/health. Use Mcp-Session-Id header for sessions.

  • Where can I get help or documentation?

Full MCP docs and client setup guides: https://app.agenticbi.com/docs/mcp-server.html and per-client docs linked there. Support: support@agenticbi.com.