HiveFlow MCP Server for AI Assistants

Overview

The HiveFlow MCP Server is an official Model Context Protocol (MCP) adapter that lets AI assistants (for example, Claude, Cursor, and other MCP-compatible agents) interact directly with your HiveFlow automation platform. It implements the MCP server side of the bridge, exposing HiveFlow flows, executions, and configuration as tools and resources the assistant can call in real time. This enables assistants to query pipeline state, create or run automations, and share context back to HiveFlow securely.

Designed for developers integrating LLM-powered assistants into automation workflows, the MCP server provides a minimal, secure shim between an assistant and the HiveFlow API. It supports local development and self-hosted HiveFlow instances, authenticates requests with API keys, and maps MCP tool calls to HiveFlow REST endpoints so assistants can act on your behalf without manual API wiring.

Features

• MCP-compatible server process for connecting assistants to HiveFlow
• Tool primitives for flow lifecycle: create, list, get, execute, pause, resume
• Access to execution history and flow metadata
• Configurable API endpoint for cloud or local HiveFlow instances
• Secure API key authentication and HTTPS communication
• Debug logging support for development and troubleshooting
• Small, dependency-light Node.js distribution (installable via npm)

Installation / Configuration

Install globally with npm:

npm install -g @hiveflow/mcp-server

Configure your MCP client (example: Cursor) by adding an MCP server entry. Example .cursor/mcp.json:

{
  "mcpServers": {
    "hiveflow": {
      "command": "npx",
      "args": ["-y", "@hiveflow/mcp-server"],
      "env": {
        "HIVEFLOW_API_KEY": "your-api-key-here",
        "HIVEFLOW_API_URL": "https://api.hiveflow.ai"
      }
    }
  }
}

For local development against a self-hosted HiveFlow backend:

{
  "mcpServers": {
    "hiveflow": {
      "command": "npx",
      "args": ["-y", "@hiveflow/mcp-server"],
      "env": {
        "HIVEFLOW_API_KEY": "dev-api-key",
        "HIVEFLOW_API_URL": "http://localhost:5000"
      }
    }
  }
}

Command-line alternative:

hiveflow-mcp --api-key YOUR_KEY --api-url https://your-instance.com

Environment variables summary:

Available Tools

The MCP server exposes a set of tools assistants can call. Each maps to HiveFlow API actions:

• create_flow — Create a new automation flow (name, triggers, steps)
• list_flows — Return a list of flows visible to the API key
• get_flow — Retrieve flow configuration and metadata by ID
• execute_flow — Start an execution with optional input payload
• pause_flow — Pause an active flow
• resume_flow — Resume a paused flow
• get_flow_executions — Fetch execution history for a flow
• list_mcp_servers — List registered MCP server configurations
• create_mcp_server — Register a new MCP server entry

Example tool call (abstract):

{
  "tool": "execute_flow",
  "params": {
    "flowId": "abc123",
    "inputs": {"email": "[email protected]"}
  }
}

Available Resources

Assistants can also access read-only resources mapped to HiveFlow data URIs:

• hiveflow://flows — All flow definitions
• hiveflow://executions — Execution records and logs
• hiveflow://mcp-servers — MCP server configuration entries

These resources let the assistant include structured context (flow schemas, recent runs) when making decisions or composing follow-ups.

Use Cases

• Create an “Email Processor” flow by prompting your assistant: “Create a flow called ‘Email Processor’ that extracts sender, subject, and routes to support.”
• Ask the assistant to list active automations: “Show me all my active flows.” The assistant calls list_flows and returns a summarized table.
• Execute a flow with runtime data: “Run flow abc123 with input { ‘email’: ‘[email protected]’ }.” The assistant issues execute_flow and reports the new execution ID and status.
• Investigate recent failures: “What failed in the last 24 hours?” The assistant queries hiveflow://executions and filters error states, returning logs or suggestions.
• Admin workflows: Register new MCP endpoints or rotate API keys through create_mcp_server and configuration management flows.

Debugging & Troubleshooting

• “HIVEFLOW_API_KEY is required”: Ensure the env var is set in your MCP configuration and that the key is valid.
• “Cannot connect to HiveFlow API”: Verify HIVEFLOW_API_URL, ensure the service is reachable, and confirm no firewall rules block the connection.
• Enable verbose logs for development:

export DEBUG=hiveflow-mcp:*

Links and Resources

• GitHub: https://github.com/hiveflowai/hiveflow-mcp-server
• HiveFlow docs: https://doc.hiveflow.ai
• MCP spec: https://modelcontextprotocol.io

This MCP server is intended for developers building integrated assistant-driven automation. Use it to give your assistant secure, live access to HiveFlow assets and execution capabilities without exposing your backend credentials directly to the agent.