📘 Facebook + Instagram
🔔
NOTIFICATIONS
No active alerts ✅
Sign Out

1. How SocialConductor.AI Works

SocialConductor.AI connects your Facebook Pages to Google's Gemini AI model to provide real-time, context-aware automated engagement — on public comments and Messenger DMs — 24/7.

End-to-End Flow

📡
Listen
Facebook sends a webhook event
🛡️
Validate
6 security gates filter
🧠
Analyze
Read post + 8 recent comments
Generate
Gemini writes a reply
Delay
15–60s human-like pause
📢
Publish
Reply posted via Graph API

Key Concepts

ConceptWhat It Means
Tenant / PageEach Facebook Page you connect is a "Tenant" with its own AI prompt, schedule, and usage limits.
System PromptThe AI's personality script. Defines tone, rules, and topics for your specific business.
Simulation ModeAI generates and saves draft replies to Logs — but does NOT post publicly. Safe for testing.
Leads VaultComments received outside business hours are saved and auto-processed when the schedule reopens.
Social ListeningThe AI reads the original post and recent comments to understand conversation context.
Activity LogA full record of every comment and AI response, searchable and exportable.

2. Trial Period & Subscription Plans

Your 14-Day Free Trial

Every new account starts with a 14-day free trial — no credit card required. Here is exactly what happens at each stage:

PeriodStatusWhat Works
Days 0–13 🎉 Trial Active Full access — live AI replies, both channels, all features at Free plan limits.
Days 14–30 ⚠️ Simulation Only AI generates log entries but does NOT post publicly. No plan upgrade = simulation mode forced ON.
Day 31+ 🔴 Polling Disabled All webhook polling is stopped. Messenger and Comments are disabled until a paid plan is activated.
Any time 💳 Paid Plan Active Full unrestricted access based on your chosen plan tier (Basic / Enhanced / Enterprise).
💡 Trial Reset: If you upgrade during Days 14–30, live posting is immediately restored. Your data and logs are never deleted — only polling is paused.

Subscription Plans

Free
$0
15 replies/day
5 image uploads
Basic analytics
Basic
$29/mo
30 replies/day
10 image uploads
Full analytics
Enhanced ⭐
$59/mo
60 replies/day
15 image uploads
Advanced features
Enterprise
$149/mo
80 replies/day
30 image uploads
Priority support

Plans are per-page per-month. Select pages using checkboxes on the Dashboard and use PayPal checkout. Hard daily safety cap: 1,000 replies per page, regardless of plan — this protects your account from Facebook velocity bans.

3. Setup Wizard

New users are automatically redirected to the 6-step Setup Wizard after their first Facebook login. You can re-run it at any time from the sidebar.

StepWhat You DoWhy It Matters
1 — Accept TermsReview and accept the EULA & Privacy PolicyRequired for GDPR compliance and to enable live posting
2 — System PromptWrite or auto-generate your AI's personalityControls every reply the AI will ever make
3 — Operating HoursSet active days and hoursEnsures AI only responds during business hours; Vault captures off-hours leads
4 — Enable ChannelsToggle Comments and/or MessengerChoose which interactions the AI handles
5 — Analytics TourLearn to read Stats and LogsUnderstand your data for informed decisions
After completing the wizard, all settings can be changed anytime via ⚙️ Edit AI on any page card in the Dashboard.

4. The System Prompt

The system prompt is the most important setting in the app. It defines who your AI is and how it speaks. The AI reads this prompt before every single reply — it is never visible to your customers.

What to Include

  • Business identity: Your business name, industry, and what you sell or do.
  • Tone of voice: Friendly, professional, concise, enthusiastic, formal, etc.
  • Key information: Hours, location, contact details, website URL.
  • Response rules: "Always end with our phone number" or "Never discuss competitor pricing".
  • Call to action: "When someone asks about orders, direct them to example.com/order".
  • Length limit: "Keep all responses under 2 sentences" or "Use bullet points for FAQs".

Example Prompts

🍕 Restaurant Example
You are "Mario's Trattoria Assistant", a warm and welcoming helper for our Italian restaurant in Chicago. We serve authentic pasta, pizza, and seafood. Our hours are Mon–Sat 11am–10pm, Sun 12pm–8pm. For reservations, always mention our website (www.mariostrattoria.com) or phone (312-555-0199). When customers ask about specials, mention that our chef posts daily specials on our Facebook page. Keep responses under 3 sentences and always end with "Buon appetito! 🍝"
💅 Beauty Salon Example
You are the AI assistant for "Glow Studio" — a boutique hair and nail salon in Austin, TX. You are friendly, upbeat, and conversational. Our services include haircuts, coloring, manicures, and facials starting at $35. For bookings use our online scheduler at glowstudio.com/book. If asked about pricing, always say prices "start from" rather than quoting exact totals. Never promise specific appointment times — always direct people to book online. End every reply with a beauty-related emoji.
🏋️ Fitness Studio Example
You are the virtual assistant for "IronCore Gym". You are motivational, direct, and energetic. We offer personal training, group classes (HIIT, yoga, spin), and open gym 24/7. Monthly memberships start at $29. For new member inquiries, direct them to our "Free 7-Day Pass" offer at ironcoregym.com/trial. When someone asks about classes, tell them our schedule is posted every Monday. If someone mentions an injury or medical concern, always recommend they consult a doctor first. Keep responses to 2–3 sentences max.
⚠️ What NOT to Include: Passwords, API keys, personal phone numbers you don't want public, competitor names you don't want the AI to mention, or anything that would be embarrassing if screenshotted and shared publicly.

5. Operating Schedule

The schedule controls when the AI processes incoming comments. This is critical for businesses that want to maintain a "human feel" — a comment at 3am getting an instant reply can feel unnatural.

Schedule Modes

ModeBehaviorBest For
Fixed Replies only within your exact set hours (e.g., 9am–5pm) B2B companies, law firms, medical offices
Random / Intermittent Replies within your set hours but adds ±30 minute variance to response timing Restaurants, retail, any brand wanting natural human-like behavior
Disabled AI replies 24/7 with no schedule restriction E-commerce, global brands, 24/7 support pages

The Leads Vault

When a comment arrives outside your schedule, it is saved to the Leads Vault (log type: vault_captured). The recovery worker (runs every 20 minutes) checks if your schedule has reopened and automatically processes any vaulted comments — so no lead is ever permanently lost.

✅ Example: A customer comments on your post at 11pm (outside your 9am–5pm window). Their comment is saved. At 9:01am the next day, the AI automatically replies to their comment as if it just saw it.

6. Channels: Comments & Messenger

Comment Reply

When enabled, the AI monitors your Facebook Page's public posts for new comments and replies automatically. The AI reads the full thread context (original post + recent comments) before generating a response.

  • Reply Threading: The AI can have multi-turn conversations. If a user replies to your AI's comment, the bot will continue the conversation thread.
  • Self-Reply Protection: The AI will never reply to comments made by your own Page admin account.
  • Conversation Detection: If a user continues a conversation (replying to the AI), the 60-minute deduplication window is bypassed to allow natural back-and-forth.

Messenger Reply

When enabled, the AI handles incoming Messenger DMs. It also captures:

  • Postbacks / Button Clicks: Logged when users tap buttons on your Page's quick replies.
  • Read Receipts: Logged when users read AI-sent messages (visible in Logs as message_read).
Facebook Note: Messenger functionality requires your Facebook app to have the pages_messaging permission and the Page to be live (not in Development mode).

7. Simulation Mode

Simulation Mode is a critical safety feature. When enabled, the full AI pipeline runs — comment detected, prompt processed, response generated — but the reply is saved to your Logs only and never posted to Facebook.

When to Use Simulation Mode

  • When you first set up a new page and want to test your prompt.
  • After making major changes to your system prompt.
  • Before a product launch or campaign, to preview AI replies.
  • If you have concerns about a specific type of comment (test with a friend's account).

Publishing Simulated Replies

In the Logs view, simulation entries have a 🚀 Post Now button. Click it to manually publish the AI's draft reply to Facebook — giving you editorial control over each response.

🔴 Auto-Forced: During the Days 14–30 trial grace period (no paid plan), Simulation Mode is automatically forced ON and cannot be disabled. Upgrade your plan to restore live posting.

8. Media Manager

The Media Manager (/images) is your image library and social publishing scheduler. It integrates directly with the Facebook Graph API to post images to your Pages.

Upload & AI Caption

When you upload an image, the system sends it to Gemini Vision. Within seconds you receive 3 ready-to-post captions:

  • 😂 Funny — Casual, lighthearted tone with humor
  • 👔 Professional — Formal, brand-focused copy with hashtags
  • 🔥 Viral — High-energy caption optimized for engagement

Visual Editor

Click 🎨 Edit Image to open the in-browser image editor. Features include: crop, resize, rotate, flip, apply filters (brightness, contrast, blur), add text overlays, and draw annotations. Edits are saved as a new copy — the original is preserved.

Scheduling

Select images using checkboxes, then use the bulk action bar to:

  • Schedule to Specific Date: Choose exact date/time for publication.
  • Random Future: System picks a random time in the near future to spread posts naturally.
  • Post Now: Immediately publish to the selected Page.

The scheduler runs every 5 minutes. Published posts show a 🔗 View button linking directly to Facebook.

Storage Limits

PlanDaily UploadsStorage
Free5 images/day50MB total
Basic10 images/day50MB total
Enhanced15 images/day50MB total
Enterprise30 images/day50MB total

📷 Shooting on a real camera? You can upload straight from a Canon, Nikon, or Sony body (including RAW) over a private SFTP connection — see Section 26 — Camera Direct Upload.

9. Safety Gate Pipeline

Every incoming comment or message passes through a sequential gate pipeline before the AI is allowed to respond. Gates run in a fixed order; the first one that fails ends processing for that event. Most blocks are logged to your Comment Logs with a reason badge (so you can see exactly why something was skipped), and a few sensitive categories are routed to human review instead of being auto-answered. This is what keeps your account safe, your costs predictable, and your replies appropriate.

The gates evolved well beyond the original six. Numbering uses decimals (0.1, 0.3, 2.5…) because new gates were inserted between existing ones over time — the numbers reflect order in the pipeline, not importance.

Stage 0 — Identity, spam & idempotency

GATE 0 — Self-Reply Protection

Drops events where the commenter is your own Page/admin account, preventing infinite reply loops.

GATE 0.0 — Away / Hold Mode

When Vacation/Away Mode is on, events are silently dropped — or, if you chose "draft silently," queued as drafts for later review instead of being posted.

GATE 0.1 — Blocked User / Personal List

Checks the commenter against your block list (dropped, logged) and your personal list (handled per your personal-list rules rather than the normal AI flow).

GATE 0.3 — Velocity Spam Detection

Always on. Blocks three patterns: velocity (too many comments from one user in 10 minutes), duplicate (the same text repeated), and template fishing (many commenters posting an identical copy-paste template to farm replies). Each is logged with its own reason.

GATE 0.2 — Auto Spam / Phishing Block

Auto-blocks comments matching known scam/phishing signatures (fake giveaways, crypto, impersonation links).

GATE 0.5 — Strict Idempotency

Ensures a given comment/DM is only processed once, even if Facebook delivers the webhook multiple times or two Gunicorn workers pick it up simultaneously. Critical for preventing duplicate replies.

GATE 0.6 — Post / Video Blacklist

If you've reserved a specific post or video for manual replies only, the AI stays out of its comments.

GATE 0.7 — Manual Reply Detection

If you've already replied to a thread by hand, the AI backs off so it never talks over you.

Stage 1 — Conversation awareness

GATE 1 — Smart Deduplication

Scans recent activity (about the last hour) for this user on this page. If we replied recently, we skip — unless the user is replying to us, in which case we allow a natural back-and-forth.

GATE 1.5 — Conversation Depth Limit

Caps how many times the AI will go back and forth in a single thread (see Section 19) so conversations don't spiral.

Stage 2 — Content & intent screening

GATE 2 — Smart Short-Comment Filter

Skips ultra-low-effort comments ("Cool!", "😍") and attachment-only comments with no text — they don't warrant a business reply and would waste credits. Genuine leads are exempt.

GATE 2.1 — Cross-User Tagging

If a commenter is clearly talking to another tagged person rather than to your page, the AI stays out of the conversation.

GATE 2.5 — Directed Hostility

Detects comments attacking the page/creator directly (insults, "you're a bot," etc.). Never auto-replied — routed to human review so you decide how to respond.

GATE 2.6 — Legal / Liability / Crisis

Flags legal threats, copyright/DMCA claims, "I reported you," and breaking-incident language. Routed to human review — the AI must never improvise on legal or liability matters.

GATE 2.7 — Grief / Medical / Sensitivity

Catches bereavement, serious illness, disability, minors disclosing their age, and mental-health crisis language. Routed to human review so a person handles anything sensitive with care.

GATE 2.8 — Accuracy-Required

When a commenter challenges a fact, date, number, or asks for a source, the reply is routed to human review rather than letting the AI guess.

Stage 3 — Human emulation, schedule & limits

GATE 3 — Human Emulation (Random Skip)

Randomly skips a share of eligible comments so your page never replies to literally everything like an obvious bot.

GATE X — Schedule Check + Leads Vault

Enforces your Operating Schedule in your configured timezone. Outside active hours, high-intent comments are still captured to the Leads Vault (Section 10) even though no reply is sent.

GATE 3 (Part 2) — Typing Delay

Adds a randomized human-like delay (roughly 30–120s) before posting, so replies don't land instantly. Social context is re-read after the delay so the AI sees the freshest state of the thread.

GATE 5 — Usage Limits

Verifies the page still has daily reply credits and monthly token budget before any AI call. Protects against velocity bans and runaway API costs.

Stage 4 — Output safety

GATE 4 — AI Content Filter

The AI reviews its own draft. If it emits the sentinel NO_REPLY or the draft fails the content check, nothing is posted. Drafts that hit a watch phrase are held (🔔) for your approval with Send / Edit buttons and an email alert, instead of auto-posting.

CIRCUIT BREAKER — Hard Daily Cap

A page-level hard cap runs ahead of the pipeline. Once a page hits its daily reply ceiling, the bot stops and emails you — a last line of defense against velocity bans and viral-post cost spikes.

10. The Leads Vault

The Leads Vault is an automatic lead preservation system for businesses that use the Operating Schedule feature.

How It Works

  1. A comment arrives while your schedule is closed (e.g., 11pm on a Saturday)
  2. Instead of ignoring it, the system saves it as a vault_captured log entry with all metadata
  3. The Recovery Worker runs every 20 minutes 24/7
  4. When the schedule opens (e.g., Monday 9am), the worker finds all vaulted entries
  5. It resubmits each entry through the full AI pipeline, which generates and posts a reply
  6. The original vault entry is deleted to prevent duplicates
The Vault stores the full comment text, the commenter's name, the post ID, and the timestamp — so the AI has full context when it processes the reply hours later.

11. Social Listening

Social Listening is the AI's ability to understand the context of a conversation — not just the individual comment.

What the AI Reads Before Replying

  • The Original Post: The text of the post that the comment appeared on (e.g., "Just launched our new summer menu! 🌞")
  • The Last 8 Comments: A sample of other recent comments on that post to understand the room's mood
  • The Specific Comment: The actual comment you are replying to

How It Affects Replies

The AI is instructed to:

  • Analyze if the room mood is: happy, celebrating, sad, technical, angry, or asking questions
  • Adapt tone accordingly (empathetic if someone is complaining, cheerful if people are celebrating)
  • Stay relevant to the post topic
  • Naturally include any affiliate links or URLs from your system prompt when contextually appropriate

12. Delays & Timing

Timing ParameterValuePurpose
Reply Typing Delay15–60 seconds (random)Mimics a human typing a response — avoids bot detection
Deduplication Window60 minutesWon't reply to the same user twice within this window (per post)
Media Scheduler RunEvery 5 minutesChecks for scheduled posts that are due to be published
Recovery Worker RunEvery 20 minutesProcesses vaulted leads + replays emergency buffer events
Bot Skip Rate20% of repliesRandom skip to appear more natural
Emergency Buffer Replay50 events/runRate-limits recovery to prevent thundering herd issues
Vault Replay per Run20 events/runPrevents overload when schedule reopens after a long holiday

13. Analytics Guide

Access via: Dashboard → any page card → 📊 Stats

KPI Cards (Top Row)

MetricWhat It Measures
Total AudienceTotal Facebook followers on the connected page (fetched live from Graph API)
Active ConversationsUnique users who engaged during the selected period, with % growth vs previous period
Open RatePercentage of Messenger messages that were read (from read receipt events)
Click-Through Rate (CTR)Percentage of interactions that included a link click or postback button press

Charts

  • Engagement Timeline: Daily interaction counts. Drag to zoom. Use the 24H/7D/30D/90D/All buttons to change the window.
  • Peak Activity Hours: Bar chart of UTC hours when your audience is most active. Use this to set optimal operating hours.
  • Top Engaged Users: Bar chart of your most active commenters/messengers.
  • Most Active Conversations: Table of posts that generated the most AI replies, with direct Facebook links.

Optimization Insight

If SocialConductor detects that your page would benefit from a plan upgrade based on activity patterns, a 🚀 Optimization Insight banner appears with a specific recommendation.

14. Reading the Comment Logs

Access via: Dashboard → any page card → 📋 Logs

The Conversation View

Each log entry is displayed as a chat bubble pair:

  • Grey bubble (left): The user's original comment or message
  • Blue bubble (right): The AI's reply. If empty, the AI was blocked by a gate or was in simulation mode.

Log Types Explained

Log TypeMeaning
commentA public comment on a post that received an AI reply
messageA Messenger DM that received an AI reply
simulationAI generated a draft reply but it was NOT posted (Simulation Mode was on)
vault_capturedComment received outside operating hours — saved for later processing
message_readA user read a Messenger message the AI sent (from Facebook read receipts)
postbackA user tapped a quick reply button on your Page

Searching & Filtering

  • Use the search bar to find logs by user name, message text, or AI reply text
  • Use time filters (24H / 7D / 30D) to narrow the view
  • Use per-page dropdown (20/50/100) to see more records at once
  • Use Export CSV to download all log data for CRM import or offline analysis

15. Alerts & Notifications

Configure alerts in ⚙️ Profile → Alert Triggers. Alerts are sent to your registered email address.

AlertWhen Triggered
Usage Warning (90%)When a page has used 90% of its daily reply limit
Circuit Breaker HitWhen a page reaches the 1,000 daily hard cap — CRITICAL
Billing AlertUpcoming plan expiry or payment issues
Login AlertA new device or browser logged into your account
Database AlertSystem automatically emails support if the DB goes offline (for SocialConductor operations team)

Email frequency can be set to: Immediate, Daily Digest, Weekly Summary, or Disabled.

16. Instagram Features

SocialConductor.AI supports both Facebook and Instagram through the same platform. Instagram is connected via your Facebook Business account or directly as an Instagram Business/Creator account.

Connecting Instagram

On the login page, use Login with Facebook if your Instagram Business account is linked to a Facebook Page (most common), or Continue with Instagram Only for standalone Instagram Business/Creator accounts. After connecting, you'll see your IG account listed under each Facebook Page it's linked to on the Dashboard.

Instagram Comment Replies

When someone comments on your Instagram posts, the AI generates and posts a reply — exactly like Facebook comments. Log entries show a 📸 Instagram badge so you can distinguish them from Facebook activity at a glance.

Instagram DMs

Instagram Direct Messages are handled automatically when DMs is toggled ON in ⚙️ Edit AI. The AI reads the incoming DM, generates a contextual reply, and sends it back via the official Messenger API. Requires instagram_manage_messages permission.

Editing Instagram Replies

Instagram does not allow editing comments via API. When you click ✏️ Edit on an Instagram log entry, SocialConductor deletes the original reply and posts a new one with your edited text. The log entry is updated with the new comment ID automatically.

Platform Filter in Logs

The Logs page has a tab bar at the top: 🌐 All, 🔵 Facebook, and 📸 Instagram. Click any tab to see only that platform's activity. The filter persists across time frame changes.

Instagram Analytics

When Instagram is linked, the Analytics page shows a dedicated Instagram Insights card with: follower count, total posts, 28-day reach, impressions, profile views, follower delta, and a recent media grid. The AI Insight Report also incorporates your Instagram data alongside Facebook metrics for cross-platform recommendations. Requires instagram_manage_insights Advanced Access.

📋 Permissions required for full Instagram features:
instagram_basic — comment reading and reply posting
instagram_manage_messages — DM reading and sending (Advanced Access)
instagram_manage_insights — follower metrics, reach, impressions (Advanced Access)

17. IG Following & Personal List

You may not want the AI to automatically reply to accounts you personally follow on Instagram, or specific Facebook commenters you've marked as personal contacts. The Personal List and IG Following Sync give you granular control.

How It Works

When someone on the Personal List or IG Following list comments, the AI still generates a reply — but holds it as a draft instead of posting. You'll see it in the Logs with a 📸 IG Following — DRAFT HELD or 🙋 Personal — DRAFT HELD badge, plus a "AI would have replied: …" preview. You can post the draft with one click or write your own reply.

IG Following Auto-Sync

Meta does not expose the following list for Business/Creator accounts via the Graph API. Instead, you import it manually from Instagram's data export — a one-time process you can repeat whenever your following list changes.

How to Import Your Following List

  1. Open Instagram → tap the ☰ menu (top right) → Settings and activity
  2. Search "down" in the settings search bar
  3. Tap Download your information — this opens Meta Accounts Center
  4. Tap Create export
  5. Select your Instagram account → under Connections, select Following
  6. Set the format to JSON → submit the request
  7. Wait for the email from Instagram (usually a few minutes to hours)
  8. Download and extract the ZIP → find following.json inside the connections/followers_and_following/ folder
  9. Go to Moderation Panel📸 Instagram Following → expand ⬆️ Import from Instagram Data Export → upload the file
Re-import whenever your following list changes — existing entries are preserved and their toggle state (Hold/Auto) is never overwritten by a re-import.

Per-Account Toggle

In the Moderation Panel, each person on the Personal List or IG Following list has a toggle switch:

  • ✋ Hold (default) — AI generates a draft but holds it for your review
  • 🤖 Auto — AI replies normally, just like any other commenter

Adding to Personal List Manually

From the Logs page, click the 🙋 Personal button on any log entry to add that commenter to your personal list. Facebook users can only be added manually (Facebook's API does not provide a friends/following list for Pages).

18. Moderation Panel

Accessible from the sidebar or via the Logs page, the Moderation Panel gives you full control over who the AI responds to.

Sections

SectionWhat it controls
🚫 Blocked UsersAI ignores all comments/DMs from these users silently
📸 Instagram FollowingAccounts auto-synced from your IG; AI holds drafts by default, toggleable per account
🙋 Personal List — ManualAccounts you've manually added; same draft-hold behaviour, same toggle
👁️ Hidden UsersComments from these users are processed normally but hidden from the Logs view

Search

Use the search bar at the top of the Moderation Panel to filter any list by username. The search applies to all sections simultaneously.

Pagination

Each section shows 20 entries per page. For large IG Following lists, use the page number controls at the bottom of each section to navigate.

CSV Exports

Three export buttons at the top right download each list as a CSV file:

  • ⬇️ Blocked CSV — username, reason, blocked date
  • ⬇️ IG Following CSV — username, bot reply status, date added
  • ⬇️ Personal CSV — username, bot reply status, date added

The Logs page also has an ⬇️ Export CSV button in the timeframe bar that downloads your last 5,000 interactions including all message content, timestamps, and reply types.

19. Conversation Depth Limits

By default the AI replies once per thread (depth = 1). This prevents runaway back-and-forth conversations that consume your daily reply budget and can feel spammy. You can increase the limit per platform in ⚙️ AI Settings → 🔁 Conversation Depth Limits.

How it works

PlatformSettingHow depth is counted
📘 Facebook Commentsfb_comment_depthBot replies on the same post thread
💬 Facebook Messengerfb_dm_depthBot replies to this user on this thread
📸 IG Commentsig_comment_depthBot replies to this comment/media thread
📩 IG DMsig_dm_depthBot replies to this sender today (resets daily)
IG DM depth resets every day. This prevents a situation where someone who DM'd you once is permanently locked out of getting future replies. Their daily conversation starts fresh each day.

In the Logs

When depth is hit, the entry shows a red 🔁 Gate 1.5 — Depth Limit or 🔁 Gate 5.5 — Depth Limit badge with an explanation: how many times the bot replied, what the max is, when the last reply was sent, and a link to the setting to adjust it.

Each reply counts against your plan

Every reply at any depth level runs through check_usage() and counts toward your daily plan limit. Reducing depth to 1 directly reduces plan consumption and Gemini API costs.

20. Lead Funnel

The Lead Funnel converts first-time DM senders into buyers using a ManyChat-style automated sequence — without requiring any third-party tools.

🎯 The flow: First DM → Welcome message + 3 topic buttons → User taps topic → Personalised reel link DM → User feels helped → AI conversation begins naturally

Setting it up

Go to ⚙️ AI Settings → 🎯 Lead Funnel. Enable it for Facebook, Instagram, or both. Then configure:

FieldPurpose
Welcome MessageSent on first DM. Should greet them warmly and ask which topic they want to learn about. End with a question to prompt them to tap a button.
Topic 1–3 Button LabelsShort labels (max 20 chars) shown as quick-reply buttons. E.g. "Train Photography"
Topic 1–3 Reel LinksInstagram/Facebook reel URL sent when they tap that topic button
Topic 1–3 DM MessagesSent before the reel link — 2–3 sentences of genuine value about that topic
CTA After ReelOne sentence appended after every reel link (e.g. "Let me know what you think!")

AI Generation

Click ✨ Generate with AI to have Gemini write all fields automatically. You can seed it from:

  • None — enter your niche and topic hints manually
  • 🅰️ Prompt A — uses your existing AI persona as the voice reference
  • 🅱️ Prompt B — uses your B variant if configured

Always review and edit the generated content before saving — the AI gives you a strong starting point that you personalise.

Funnel Contacts

Every person who enters the funnel is tracked in the Funnel Contacts table (expandable at the bottom of the Lead Funnel settings). Columns show their PSID (anonymised), platform, current stage (welcomedtopic_sentin_automation), which topic they chose, and last activity time.

Instagram limitation: Meta does not expose new-follower webhook events for Business accounts. The funnel triggers on first DM, not on follow — this is the same approach ManyChat uses in practice.

21. Block & Hide — Platform-Level Actions

The Block and Hide buttons in the Logs page now make real Facebook Graph API calls in addition to updating your local database.

Block User

Clicking 🚫 Block User on a log entry does two things simultaneously:

  • Adds the user to your blocked_users table — the AI ignores all future comments and DMs from them (Gate 0.1)
  • Calls POST /{page-id}/blocked with their PSID — bans them from your Facebook Page at the platform level, so they can no longer comment publicly
Requires pages_manage_engagement permission. If the API call fails, the local block still applies and a specific flash message explains the issue.

Hide Comment

Clicking 👁️ Hide Posts does two things:

  • Adds the user to hidden_users — their entries are filtered from your Logs view
  • Calls POST /{comment-id} with is_hidden: true — hides that specific comment from your Page's public view. The commenter can still see it (Facebook's standard hide behaviour) but no one else can.
Hide is per-comment (the one you clicked on), not per-user retroactively. Block is per-user permanently.

22. Bell Notifications

The 🔔 bell icon in the top-right topbar appears on every page and shows system alerts relevant to your account.

Alert types

AlertTriggerResets
🛠️ MaintenanceA system message inserted by SocialConductor adminsWhen the message is removed
⚠️ Near LimitA page reaches 90% of its daily reply planNext day (key includes date)
🚨 At LimitA page hits 100% of its daily reply plan — AI stops replyingNext day

Dismissing alerts

Click the button on any alert to dismiss it. Dismissed alerts are stored per-user in the alert_dismissals table. Usage alerts reappear the next day automatically since their key includes today's date. Maintenance alerts stay dismissed until a new message is posted.

How it renders

Alerts are rendered server-side inline in the page (no JavaScript fetch required) so there is no "Loading..." flicker and no dependency on a separate API call succeeding. The bell badge count reflects the number of active, non-dismissed alerts.

23. Security Architecture

🔐Access Token Encryption at Rest

All Facebook and Instagram access tokens are encrypted in the database using Fernet symmetric encryption (AES-128-CBC). The encryption key is stored as an environment variable and never in source code. Tokens are decrypted only at the moment of an API call and never logged.

🛡️Prompt Injection Protection

User-generated content (comments, DMs) is always passed as user content — never concatenated into the trusted system instruction. A security guardrail is appended to every AI prompt preventing it from disclosing system configuration, tokens, or user data even if instructed to by a crafted comment. Viral few-shot examples are screened for injection phrases before storage.

🔏HMAC-SHA256 Webhook Verification

Every incoming Facebook webhook is verified using a secret-key HMAC signature. Any event that fails this check is rejected with HTTP 403.

🕵️CSRF Token Protection

All form submissions and API calls use unique per-session CSRF tokens combined with referrer validation. This prevents cross-site request forgery attacks on critical actions.

🚦Rate Limiting

The webhook endpoint is rate-limited to 100 requests per 60 seconds per IP. All API command endpoints (including OpenClaw) are rate-limited per user/key. Brute-force protection includes delays on invalid coupon attempts.

🗄️SQL Injection Prevention

All database queries use parameterized queries via psycopg2. No raw string interpolation is used for SQL. All user inputs are sanitized and length-limited before storage.

Privacy Policy Data Deletion Request Data Processing Addendum

24. Billing & PayPal Integration

How Payments Work

  1. Select one or more pages using checkboxes on the Dashboard
  2. A yellow billing bar appears — choose your plan (Basic/Enhanced/Enterprise) and billing type (30 Days or Recurring)
  3. The total is calculated server-side (not modifiable by the client)
  4. Click the PayPal button and complete payment in the PayPal pop-up
  5. On success, the selected pages are instantly upgraded in the database
  6. A subscription expiry date is set 30 days from payment

Automatic Downgrade

When you log in to the dashboard, the system checks all your pages for expired subscriptions. Any page where subscription_end has passed is automatically downgraded to the Free plan. You will see a notification in the dashboard.

Coupon Codes

Promo codes can be redeemed in the "Redeem Coupon" bar on the Dashboard. Each code adds a fixed number of bonus credits to a selected page. Bonus credits are used as an overdraft once the daily plan limit is reached. Each coupon is single-use.

25. Troubleshooting

❓ The AI isn't replying to comments
Check these in order:
1. Comment Reply toggle — Is it ON in ⚙️ Edit AI?
2. Simulation Mode — Is it OFF? If ON, replies go to Logs only.
3. EULA — Have you accepted the Terms of Service? Check Profile page.
4. Operating Schedule — Is it enabled but the current time is outside your set hours?
5. Gate 2 — Was the comment under 4 words? ("Cool!" won't get a reply)
6. Gate 1 — Did you comment yourself with an admin account? (Self-reply protection)
7. Daily Limit — Has the page hit its daily reply limit? Check Dashboard stats.
8. Trial Status — Is your trial expired (Day 14+)? Upgrade or check the banner.
❓ "Post Now" button shows an error on the Logs page
This usually means "Original comment likely deleted". The user removed their comment before you clicked Post Now. Facebook's API cannot reply to deleted content. The log entry is preserved for your records but the comment no longer exists publicly.
❓ Logs page shows no activity
1. Ensure you're on the correct page (select from Dashboard).
2. Try the time filter "All" to see historical data.
3. If you recently reconnected your Facebook account, log out and back in — your access token may have been refreshed and the webhook needs re-registration.
4. Check that your Facebook Page is subscribed to the webhook (done automatically on login, but can be re-triggered by reconnecting).
❓ Analytics charts show no data
The analytics are built from activity logs. If you have no logs yet, charts will be empty. Give the AI time to accumulate data — usually 24–48 hours of activity is needed for meaningful charts. Check if the 24H filter is selected while the page is new — try All to see all historical data.
❓ Image editor is slow to load or shows a blank canvas
The image editor loads the image as a binary blob to bypass browser security restrictions. On slow connections this can take 3–5 seconds. If the canvas remains blank after 10 seconds: refresh the page, try a different browser, or ensure you are not using a VPN that might be blocking the media asset endpoint.
❓ Facebook OAuth shows "This app is in development mode"
Your Facebook Developer App needs to be submitted for review and moved to Live mode before non-admin users can connect. Go to developers.facebook.com → your app → App Review → set the App Mode to Live. Note that Live mode requires you to have completed the privacy policy URL and other required fields.
❓ PayPal payment succeeded but plan was not upgraded
This is very rare (usually a network timeout during the capture). Contact [email protected] with your PayPal transaction ID and we will manually apply the upgrade within 24 hours. To prevent this in future, ensure your browser does not close the PayPal pop-up before the "Payment Successful" confirmation appears.
📧 Still need help?

Email [email protected] or call 571-758-2754. Include your Page ID, a description of the issue, and screenshots if available. Typical response time: 24–48 hours (Mon–Fri).

26. Camera Direct Upload (SFTP)

Camera Direct Upload lets you push photos straight from a camera body into your Media library over a private, encrypted SFTP connection. Modern Canon, Nikon, and Sony cameras can FTP images as you shoot — this feature gives each account its own isolated SFTP login, auto-develops the images (including RAW), and drops them into an approval queue. It was built for event, airshow, and trackside photographers who want shots online minutes after pressing the shutter.

How it works

  1. Provision a session. In the Camera panel, name your shoot and click Provision. You receive a dedicated SFTP host, port, username, and password. The password is shown once — store it in your camera's FTP settings right away.
  2. Configure your camera. Enter the host/port/username/password in your camera's FTP/SFTP transfer settings. Passwords are alphanumeric-only for compatibility with camera virtual keyboards (Canon 1DX/R-series reject special characters).
  3. Shoot. As images transfer, the server picks them up automatically, develops RAW files, applies adaptive enhancement, and stores the result.
  4. Review & approve. Uploaded shots land in the Ingest queue. Approve the keepers (they move into your Media library and become postable like any other image) or reject the rest.

Image processing

Every upload is processed entirely in memory and stored exactly like a web upload, so it works with scheduling, captions, and posting with no extra steps. When AI editing is enabled, the pipeline applies (in order): RAW development, gentle denoise, shadow lift, auto-levels, and capped sharpness/saturation/contrast, then resizes to 4096px on the long edge and saves at high-quality JPEG. With AI editing off, RAW files are still developed but otherwise left untouched.

SupportedFormats
JPEG.jpg / .jpeg
Canon RAW.cr3 / .cr2 / .crw
Nikon RAW.nef / .nrw
Sony RAW.arw / .srf
Adobe DNG.dng

Security & isolation

  • Each account gets its own chroot-jailed SFTP user — you can only ever see your own upload directory.
  • SFTP passwords are stored encrypted at rest and can be rotated at any time from the Camera panel (rotating also re-creates the login if it was lost during a server rebuild).
  • Every state change (provision, rotate, file received, ingest, approve, reject) is written to an audit event log.
  • Upload paths are validated to block path-traversal attempts before any file is processed.
One login per account. If you try to provision a second time, you'll get an "already provisioned" response — use Rotate Credentials instead of creating a new login. Rotate immediately if a password is ever exposed.

27. OpenClaw & Claude Code

OpenClaw is the integration layer that lets you control your SocialConductor bot from outside the web dashboard — from a chat app, the Claude Code / OpenClaw skill, or the SocialConductor iPhone app. It exposes a secure, token-authenticated API so you can check status, read logs, manage leads, post and regenerate replies, edit prompts, and pull analytics without opening a browser.

Linking your account

  1. Request a one-time link from your chat app or the mobile app. This generates a short-lived link token.
  2. Open the link (/openclaw/link). It validates the token and runs the normal Facebook OAuth handoff, tying your OpenClaw access to your real page.
  3. Once linked, an API key (Bearer token) is issued. All subsequent commands authenticate with that token — there's no password to share.

Link tokens expire quickly by design. If a link says it's expired, just request a fresh one.

The Magic Prompt

Rather than writing a system prompt by hand, the Magic Prompt generator reads your real Facebook Page profile (bio, about, description, categories) through the Graph API and drafts a tailored prompt for you. You can nudge it with optional voice, audience, and avoid hints. The page profile is cached for 7 days to keep it fast. This is the same engine behind the web "auto-generate prompt" button — see Section 4.

What you can do over OpenClaw

  • Monitor: live status, page list, comment logs, drafts, leads, blocked users, viral vault, security audits.
  • Act: post a manual reply, regenerate an AI reply, approve drafts in bulk, hold/release replies, block/unblock users, toggle vacation mode.
  • Configure: read and update your system prompt, advanced settings, and live settings.
  • Analyze: standard and enhanced analytics, A/B stats, sentiment trends, commenter profiles, top videos, and AI insights.
  • Teach: feed examples back to the AI to shape its voice over time.
  • Reviews: read, draft, generate, and post Business Review replies.

Security model

  • OpenClaw endpoints use Bearer-token auth and are rate-limited per key (for example, Magic Prompt generation is capped per page per day).
  • The token is scoped to your linked page(s) only — strict tenant isolation applies exactly as it does on the web.
  • OpenClaw routes are exempt from browser CSRF (they carry no cookie) because the Bearer token itself is the credential; they are never reachable with a stale session cookie.

28. iPhone App

The SocialConductor iPhone app puts your comment automation in your pocket. It's available on the App Store and connects to the same backend as the web dashboard, so anything you do in the app is instantly reflected on the web and vice-versa.

Download on the App Store →

Signing in

The app uses the same OpenClaw link + Facebook OAuth handoff described in Section 27. After you authorize, the app stores a Bearer token securely in the iOS Keychain — never in plaintext — and uses it to reuse the same secure API the web dashboard uses behind the scenes.

What's in the app today

  • Unified chat view across platforms with date separators and per-platform tabs.
  • Status & control: pause/resume replies, hold drafts, toggle vacation mode.
  • Comment logs with the comment's context and vibe, and one-tap AI-reply regeneration in the reply editor.
  • Leads & moderation: review captured leads, block/unblock users, post manual replies.
  • Magic Prompt generation for fast onboarding.

Platform coverage & roadmap

Facebook & Instagram are wired up first. TikTok and YouTube control from the app, full analytics views (learning analytics, top videos), per-platform settings screens, and push notifications for new replies across all platforms are on the near-term roadmap. Until push lands, use the in-app refresh and the email alerts described in Section 15 to stay on top of activity.

TestFlight / iOS note: session cookies don't survive the OAuth bounce between the LAN backend and the public domain, which is why the app authenticates with a Bearer token carried in the OAuth state rather than a browser cookie. If a sign-in ever stalls, fully close and reopen the app to request a fresh link.

29. Contact & Support

Reach the SocialConductor.AI team directly. Typical response time is 24–48 hours, Monday–Friday.

Sales [email protected]
Plans, agency/multi-account setups, and upgrades.
Support [email protected]
Technical issues, billing questions, and account help. Include your Page ID and screenshots.
Phone 571-758-2754
Mon–Fri. Email is fastest for issues that need logs or screenshots.

For account-specific requests, email from the address on your account and include your Page ID so we can verify and resolve faster.