RelayRailRelayRail
Documentation

LangChain

Add human-in-the-loop capabilities to your LangChain agents.

Prerequisites

  • A LangChain project (Python or JavaScript)
  • A RelayRail account with an API key
  • MCP support in your agent framework

Integration Approach

LangChain can integrate with RelayRail in two ways:

1. MCP Tool Integration

If your LangChain setup supports MCP, add RelayRail as an MCP server:

# Start the RelayRail MCP server
npx @relayrail/server start

# Configure your LangChain agent to use MCP tools

2. HTTP API Integration

For custom integrations, use the RelayRail HTTP API directly:

import requests

def request_approval(message: str) -> dict:
    response = requests.post(
        "https://relayrail.dev/api/mcp/request-approval",
        headers={
            "Authorization": f"Bearer {RELAYRAIL_API_KEY}"
        },
        json={
            "message": message,
            "type": "yes_no"
        }
    )
    return response.json()

LangChain Tool Example

Create a custom LangChain tool for approvals:

from langchain.tools import BaseTool
from typing import Optional

class HumanApprovalTool(BaseTool):
    name = "human_approval"
    description = "Request human approval via SMS/email"

    def _run(self, message: str) -> str:
        result = request_approval(message)
        if result.get("approved"):
            return "Approved"
        return "Denied"

# Add to your agent
tools = [HumanApprovalTool(), ...]
agent = create_agent(llm, tools)

Use Cases

  • -ReAct agents - Add approval steps between reasoning and action
  • -Autonomous agents - Gate critical actions with human oversight
  • -Data pipelines - Notify on completion or errors