Cookbook

Three task-oriented recipes you can copy and run today — no SociaHive UI required. Each one is a real automation built through the same MCP tool surface an AI assistant or a script calls: auto-reply to Instagram comments with a DM, bulk-schedule a week of posts from a CSV via the REST API or CLI, and pull a daily performance recap into your AI assistant's chat. Pick the one that matches your workflow below.

Auto-reply to Instagram comments mentioning "price"

Persona: AI user (Claude.ai, Claude Desktop, Cursor) Time: 2 minutes Code: none — pure prompt via MCP

What you'll build: a SociaHive automation flow that DMs anyone who comments "price", "cost", or "how much" on your Instagram posts. You do it by talking to your AI assistant — no SociaHive UI clicking.

Prerequisites: an MCP client connected to SociaHive — see the MCP setup guide.

The prompt:

Using SociaHive, build me an Instagram comment auto-reply flow:

1. Trigger on comments containing "price", "cost", or "how much"
2. Reply publicly with "DMing you now! 📩"
3. Send the commenter a DM with our pricing link: https://example.com/pricing
4. Activate the flow when done

Use my first Instagram account. Show me the final flow before activating.

What the AI does under the hood:

1. list_accounts                  → find your Instagram account
2. generate_flow (0 AI credits)   → AI builds the entire flow graph
3. get_flow                       → show you the result
4. activate_flow                  → after you confirm

Variations:

Cost: 0 AI credits — generate_flow is creation AI and is free on every plan. Future auto-replies are also free (template messages cost nothing; AI-personalized replies cost 5 credits each).

Schedule a week of posts from a CSV

Persona: Developer Time: 5 minutes Code: Python + Node

Prerequisites: a SociaHive API key (see the developer quickstart) and either the Python or Node SDK installed.

CSV format:

text,platform,scheduled_at,image_url
"Monday motivation 💪",instagram,2026-06-01T09:00:00Z,https://cdn.example.com/mon.jpg
"Tuesday tip: keep it short",twitter,2026-06-02T09:00:00Z,
"Wednesday wisdom",linkedin,2026-06-03T09:00:00Z,

import csv
import os
import sys
from sociahive import SociaHive
 
sh = SociaHive(api_key=os.environ["SOCIAHIVE_API_KEY"])
 
with open(sys.argv[1]) as f:
    rows = list(csv.DictReader(f))
 
account_cache = {}
for platform in {r["platform"] for r in rows}:
    accounts = sh.accounts.list(platform=platform).get("data", [])
    if not accounts:
        sys.exit(f"No connected account for {platform}")
    account_cache[platform] = accounts[0]["id"]
 
payloads = [{
    "text": r["text"],
    "platforms": [{"platform": r["platform"], "account_id": account_cache[r["platform"]]}],
    "schedule_type": "scheduled",
    "scheduled_at": r["scheduled_at"],
    "timezone": "UTC",
    **({"media": [{"type": "image", "url": r["image_url"]}]} if r["image_url"] else {}),
} for r in rows]
 
# Use bulk endpoint when >10 posts; otherwise loop
if len(payloads) > 10:
    result = sh.posts.bulk(payloads)
    for post in result["data"]:
        print(f'scheduled {post["id"]} on {post["platforms"][0]["platform"]}')
else:
    for p in payloads:
        post = sh.posts.create(**p)
        print(f'scheduled {post["id"]} on {p["platforms"][0]["platform"]}')

CLI one-shot (no scripting needed):

sociahive posts create \
  --text "Monday motivation 💪" \
  --platforms '[{"platform":"instagram","account_id":"…"}]' \
  --schedule-type scheduled \
  --scheduled-at "2026-06-01T09:00:00Z" \
  --media '[{"type":"image","url":"https://cdn.example.com/mon.jpg"}]'

cURL — single call:

curl -X POST https://www.sociahive.com/api/v1/posts \
  -H "X-API-Key: $SOCIAHIVE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "text": "Monday motivation 💪",
    "platforms": [{"platform":"instagram","account_id":"…"}],
    "schedule_type": "scheduled",
    "scheduled_at": "2026-06-01T09:00:00Z"
  }'

Error tips:

• 413 on /posts/bulk — batch cap is 100 rows; chunk your CSV.
• 400 platform_user_id required — account-id lookup matched the wrong account; pass platform_user_id explicitly.
• 429 — default 1000 req/hr per key; cache account IDs and use the bulk endpoint.

Verify: sociahive posts list --status scheduled --limit 20.

Cost: $0 in AI credits.

Daily performance recap with Claude

Persona: Power user / agency operator Time: 2 minutes Code: none — pure prompt via MCP

What you'll get: a one-screen morning brief — which flows underperformed, which posts overperformed, what's queued for today, and where attention is needed. No clicks, no SQL, no dashboard hunting.

Prerequisites: an MCP client connected to SociaHive — see the MCP setup guide.

The prompt:

Give me a brief on yesterday's SociaHive performance. Steps:

1. List my flows with stats for the last 24 hours.
2. Pull analytics overview for the same window across all platforms.
3. List anything scheduled to publish today.
4. Output a 6-line summary:
   • Top performing flow (by completions) and why
   • Worst performing flow (by error rate) and what to investigate
   • Top post (by engagement) and what it had in common
   • Total reach / impressions across platforms vs the previous day
   • Anything queued for today I should review before it ships
   • One concrete action you'd recommend

Be concise — under 200 words total. Do not modify anything.

What the AI does under the hood:

1. list_flows_with_stats     → flow performance + error counts
2. get_analytics_overview    → cross-platform impressions / engagement
3. list_scheduled_posts      → today's queue

All three are read-only tools — no risk of accidental changes, no confirmation prompts.

Variations:

Cost:

• 0 SociaHive AI credits — all three tools are read-only metrics endpoints, not metered.
• Whatever your AI assistant charges for its own tokens (usually pennies per run with Claude Haiku or GPT-4o-mini).

Why this beats the dashboard:

The dashboard shows you data. The AI tells you what to do about it — which is what you actually wanted when you opened the dashboard at 9am.

Common questions

Do I need to write code for any of these? Recipes #1 and #3 are pure prompts — you type a sentence to your AI assistant and it calls the MCP tools for you. Recipe #2 (CSV import) has a Python and Node example, but the CLI one-shot and cURL examples in that section work without writing any code at all.

Can I run these with ChatGPT instead of Claude? Yes. Every recipe here uses SociaHive's public MCP server, which any MCP-compatible agent can call — Claude Desktop, Claude Code, ChatGPT Connectors, or Cursor. See the MCP setup guide for per-client connection steps.

What happens if a step fails partway through? Reads (list_accounts, get_flow, list_flows_with_stats) are always safe to retry. Writes that are irreversible — publishing a post right now, deleting a flow — return a short-lived confirmation token on the first call and only run on a second, explicit confirm; nothing destructive fires on a single call. For the CSV recipe specifically, see the error tips above for the common 413 and 429 failure modes and how to recover.

How much do these recipes cost? Recipes #1 and #3 are $0 in AI credits — flow generation and read-only analytics tools are free creation AI, not metered. Recipe #2 doesn't call AI at all; it's direct REST/CLI calls, so it costs nothing beyond your own AI assistant's token usage (if you're driving the CLI from one).

What to read next

• Flow patterns — the seven shapes most SociaHive flows take, so you can describe what you want to build.
• Developer quickstart — REST, SDKs, and CLI reference if you want to wire SociaHive into a larger application.
• MCP setup guide — the AI-assistant integration that unlocks recipe #1 and #3.