Knowledge Base

On the dashboard click 'Connect Facebook' and authorize via the official Meta OAuth dialog. We never see your password. The flow takes about 60 seconds — you approve the scopes we listed, and the page appears under 'Social accounts' immediately. We then sync the last 7 days of comments so the bot starts with context.

Instagram Business accounts must first be linked to a Facebook Page (Meta's requirement, not ours). Once linked, our Facebook OAuth flow automatically detects the linked IG and adds it as a separate social account. Personal Instagram accounts cannot use Mochu — only Business and Creator accounts.

By default the bot starts in semi-auto mode: it drafts replies but waits for your approval. We strongly recommend keeping semi-auto on for the first week so you can validate the tone. After that, flip it to full-auto in account settings and the bot publishes replies in <2 seconds without you.

Beauty salons, restaurants, e-commerce stores (under €50k MRR), fitness studios, real estate agencies, dermatology clinics, and marketing agencies running white-label. The shared trait: paid Meta Ads budget + B2C customer base + comments under ads with purchase intent.

Zero. Configuration is three clicks: connect Meta (1 click OAuth), choose your industry preset (1 click), enable auto-moderation (1 click). The bot inherits sensible defaults for your industry. No regex, no API keys, no Zapier.

Every new comment is sent to OpenAI's GPT-4 family via our classify-comment edge function. The model returns one of 6 categories (lead / question / hate / spam / praise / other), the detected language, a confidence score and lead-intent metadata. Whole round trip takes ~1.5 seconds. The classifier is fine-tuned on 50 000+ labelled comments across our 47 supported languages.

47 languages including all EU languages (Polish, German, French, Italian, Spanish, Portuguese, Dutch, Hungarian, Czech, Slovak, Bulgarian, Croatian, Slovenian, Greek, Finnish, Swedish, Danish, Norwegian, Estonian, Latvian, Lithuanian, Maltese, Irish), Eastern (Ukrainian, Russian, Turkish), MENA (Arabic, Persian), South Asia (Hindi), East Asia (Japanese, Korean, Chinese), Southeast Asia (Vietnamese, Indonesian, Thai).

Every decision is logged in our audit log with the prompt, model name, response, latency and confidence. You can reverse any AI action (unhide a comment, delete a sent reply, undo a DM) with a single click from the audit view. Repeated misclassifications can be fixed by editing the per-account custom instructions field in account settings — the AI inherits those on every classification call.

Two-layer logic: first the classifier categorises (lead / question / hate / spam / praise / other), then your account-level auto-moderation toggles decide what action to take per category. You can enable auto-hide-hate without enabling auto-DM-leads, for example. By default new accounts have only auto-hide-spam enabled to avoid surprising the customer.

Yes — but only in two scenarios: (1) when a comment is classified as a lead AND auto-DM is enabled in account settings; (2) when the operator manually clicks 'Send DM' on a comment row. The bot never DMs proactively to users who didn't engage first. We respect Meta's 24-hour messaging window strictly.

Yes. Each account has three settings that influence every AI reply: tone (friendly / formal / casual / professional), style (long / short / emoji / no-emoji), and custom instructions (free text up to 500 chars where you describe your business and voice). The classifier and reply generator both read these on every call.

Most won't — replies are short, in your brand voice, in the customer's language, and use the customer's first name. We surveyed 312 end-customers who received a Mochu-generated reply: 84% rated the response 'helpful', 11% 'neutral', 5% 'this felt like a bot'. The 5% number drops to 1% if the operator adds 30+ chars of custom instructions explaining their business.

Starter (99 PLN/mo): 1 Facebook Page + 1 Instagram, AI classifier, auto-hide hate, auto-DM leads, lead dashboard, email support. Pro (199 PLN/mo): 5 FB + 5 IG, everything from Starter, DM Inbox with auto-replies, Meta Ads ROI monitor, weekly email report, Lead Inbox + CSV export, priority support. Agency (499 PLN/mo): 20+ accounts, everything from Pro, white-label panel, dedicated account manager, custom integrations (API/webhooks), 99.9% SLA, team onboarding.

Yes — 14 days, no credit card required. After day 14 the account moves to a free read-only mode (you can see your data but the bot stops) until you pick a paid plan. We never charge a card without explicit consent.

Yes, from the billing screen in your dashboard. Upgrades take effect immediately and we prorate the difference; downgrades take effect at the start of your next billing cycle.

Yes — pay annually and get 20% off all three tiers. This is also the recommended option if you want to lock in current prices (we reserve the right to raise prices for new monthly customers as Meta API costs change).

30-day money-back guarantee, no questions asked. If Mochu doesn't help you recover at least 5 leads in your first month, we refund 100%. After 30 days we don't refund partial months but you can downgrade or cancel anytime.

PLN, EUR, RON, HUF, CZK, BGN, DKK, SEK, NOK, RUB, UAH, TRY, JPY, KRW, INR, BRL, USD, GBP. Billing currency follows your IP/locale on signup; you can change it once from settings.

Yes. We use Meta's official OAuth — we never see your password and never can. Every access token Meta issues is encrypted at rest in a Supabase Vault. We are a registered Meta partner and have passed their app review process. If you disconnect Mochu at any time, we delete the token within 60 seconds and purge cached data within 24 hours.

Yes. We process customer data only for the purpose of providing the moderation service. We don't sell, share or use customer data for AI training. A Data Processing Agreement (DPA) is available at /pl/dpa or on request. You can export all data from your account or request deletion at any time.

Mochu runs on Supabase in the EU (Frankfurt region). All data — tokens, comments, audit logs, customer profiles — stays in EU jurisdiction. We don't replicate to non-EU regions.

Your subscription becomes inactive immediately. Read access to your data continues for 30 days so you can export. After 30 days we permanently delete: tokens, comments, audit logs, organisation settings. Account profile (email, name) is kept for an additional 60 days for fraud-prevention purposes, then deleted.

No production incidents to date. We run automated security tests on every deploy, manual security review quarterly. We accept security disclosures at security@mochu.io per our /.well-known/security.txt policy.

Macro-F1 of 94% across 6 classes (lead / question / hate / spam / praise / other) on a labelled holdout set of 50 000+ comments. For Polish-specific hate detection we hit 99.7% precision on a separately tested 12 000-comment set. We re-run the test quarterly and publish updated numbers on /statistics.

Median round-trip from Meta webhook delivery to our reply hitting Meta's Graph API is <2 seconds (P50). P99 is <9 seconds. Most of the latency budget is the OpenAI classifier call (~1.5s) — the rest is database writes and Meta's reply API.

Yes — Mochu processes 127 000+ comments daily across all customers. For an individual viral post our global rate limiter caps outbound replies at 60/minute per page so we never trip Meta's quota. If you're expecting a viral spike (e.g. major product launch) write to support — we can pre-warm capacity.

Median is <4 hours after enabling auto-DM. P75 is <24 hours. If you don't see at least one recovered lead in your first 7 days, write to hello@mochu.io and we'll audit your setup.

ManyChat is a chatbot platform for DM flows (mainly Messenger and Instagram DMs). Mochu starts with Meta Ad comments — the surface ManyChat doesn't cover. We do DMs as a follow-up upsell, not as the primary product. Pricing: ManyChat $15-$80/mo for limited flow nodes; Mochu 99-199 PLN/mo for unlimited comment + DM + auto-moderation.

Hootsuite and Sprout Social are general social-media management tools — scheduling, multi-channel inboxes, basic analytics. Their pricing starts at $99/mo and goes up steeply with seats. Mochu does one thing (Meta Ad comment recovery) and does it natively with AI. For a small business that doesn't need scheduling-as-a-service, Mochu is 5-10× cheaper.

Meta's built-in tools handle profanity filtering (keyword block lists) and basic auto-hide on flagged words. They don't (1) classify intent, (2) reply automatically with context, (3) trigger DM flows from comments. Mochu adds the missing layer.

Agency plan (499 PLN/mo) includes: custom-branded login page, your domain on dashboard URL, your logo on the in-app navigation, customisable colour scheme, custom 'powered by' attribution. Clients log in to YOUR domain — they don't know Mochu exists.

Yes. The organisation_members system supports multiple organisations per user with role-based access (owner / admin / editor / viewer). Switching between client accounts is a single dropdown click.

Agencies billed under their own brand keep 100% of what they bill their clients. Mochu charges agencies the flat Agency plan (499 PLN/mo per workspace) regardless of how many client accounts are inside. Some agencies bill clients €100/mo for Mochu and pocket the spread.

Use the bulk-invite API endpoint — POST a CSV of client emails to /api/organizations/bulk-invite and we send each one a self-onboarding link branded with your agency. Each client connects their own Meta account; the agency never touches client passwords.