The agent framework backbone, with SociaHive as a first-class tool node.

LangChain has the largest production-agent footprint of any framework — Python and JS both, with LangGraph as the go-to for stateful, branching workflows. The langchain-mcp-adapters package (real, on PyPI and npm) gives you a one-line connection to any MCP server, including SociaHive's at https://sociahive.com/api/mcp. Once connected, every SociaHive tool (create_post, generate_flow, get_post_analytics, list_accounts, …) appears as a LangChain Tool you can pass to an agent or wire as a node in a LangGraph graph.

Why pair LangChain with SociaHive

Register SociaHive in a LangChain MultiServerMCPClient

The langchain-mcp-adapters package supports MCP servers over both stdio and HTTP. SociaHive runs HTTP, so use the streamable_http transport with header auth. The agent gets every SociaHive tool as a regular LangChain Tool — pass it to AgentExecutor, LangGraph, or any tool-using primitive in the framework.

# pip install langchain langchain-mcp-adapters langchain-anthropic
from langchain_mcp_adapters.client import MultiServerMCPClient
from langgraph.prebuilt import create_react_agent
from langchain_anthropic import ChatAnthropic
import os

client = MultiServerMCPClient({
    "sociahive": {
        "transport": "streamable_http",
        "url": "https://sociahive.com/api/mcp",
        "headers": {
            "X-API-Key": os.environ["SOCIAHIVE_API_KEY"],
        },
    },
})

# Tools are discovered from SociaHive's MCP advertisement.
tools = await client.get_tools()

agent = create_react_agent(
    ChatAnthropic(model="claude-sonnet-4-6"),
    tools,
)

result = await agent.ainvoke({
    "messages": [
        ("user", "Pull last 7 days of Instagram + LinkedIn analytics, "
                 "summarise the top 3 posts, and draft a Threads recap."),
    ],
})

A real LangGraph workflow: weekly cross-platform recap with retry policy

A small LangGraph state machine pulls analytics from every SociaHive-connected platform, filters to the top performers, drafts a recap, and stages a scheduled post. Errors on any platform fan out to a fallback branch instead of failing the whole run.

1. 1Node 'fetch_analytics': calls get_post_analytics for each platform in parallel using LangGraph's parallel branches. SociaHive returns post-level data; the node aggregates by engagement rate.
2. 2Node 'filter_top': keeps the top 3 posts across all platforms. If any platform errored (rate limit, deauth), the error edge routes to 'notify_slack' instead of breaking the graph.
3. 3Node 'draft_recap': calls the LLM with the top-3 summaries and your brand voice file. Output is a Threads-length post + a longer LinkedIn variant.
4. 4Node 'stage_post': calls create_post twice — once per platform — with scheduledAt set to next Sunday 6pm. Returns SociaHive's pending_confirmation envelope for the publish step (publishing is destructive).
5. 5Node 'notify_human': posts a Slack message with the confirmation URLs so you can approve from the dashboard before SociaHive actually publishes.

# Inside the graph definition. SociaHive tools came from MultiServerMCPClient.

async def fetch_analytics(state):
    by_platform = {
        "instagram": state["accounts"]["instagram_id"],
        "linkedin":  state["accounts"]["linkedin_id"],
        "tiktok":    state["accounts"]["tiktok_id"],
    }
    results = await asyncio.gather(*[
        get_post_analytics_tool.ainvoke({
            "platform": p,
            "accountId": acc_id,
            "since": "2026-04-30",
        })
        for p, acc_id in by_platform.items()
    ], return_exceptions=True)
    return {"analytics": results}

async def stage_post(state):
    # SociaHive returns pending_confirmation here — destructive action.
    res = await create_post_tool.ainvoke({
        "platform":    "linkedin",
        "accountId":   state["accounts"]["linkedin_id"],
        "caption":     state["recap_long"],
        "scheduledAt": state["sunday_6pm_iso"],
    })
    return {"pending_confirmation": res}

# Notes:
#   - generate_flow consumes AI credits (refunded on error).
#   - create_post / schedule_post do NOT consume AI credits.
#   - publish_post_now and activate_flow return pending_confirmation
#     envelopes — surface them to a human via 'notify_human' node.

What You Can Do with LangChain + SociaHive

• ✓Connect once via langchain-mcp-adapters; every SociaHive tool flows into your existing LangChain agents
• ✓Build LangGraph state machines that span Instagram, WhatsApp, LinkedIn, Threads, TikTok and the rest
• ✓Production-grade observability — LangSmith traces every SociaHive tool call alongside your other steps
• ✓Same Python or JS code talks to SociaHive in dev and prod; no per-platform glue layer

How to Connect LangChain

Use Cases with LangChain

Other AI Agent Integrations

Frequently asked questions