A production system prompt
Not a one-liner. It names the exact tools your agent has, tells it how to behave when a send is rate-limited instead of hammering the number, states what it must never claim, and spells out its escalation path to a human.
Agent Studio · free on every tier
Answer 10 questions. Get a production-grade WhatsApp AI agent.
Agent Studio is a free WhatsApp AI agent prompt generator, built into the app. Tell it about your business and it assembles a production system prompt, a ready-to-paste n8n workflow and a Claude/MCP config — in a second, offline. No AI writes it: it fills in a hand-authored template with your answers. Which is exactly why it is free, instant, and identical every time.
In short
Agent Studio is a free WhatsApp AI agent prompt generator built into the SocialMate desktop app (sidebar → AI Agent). Answer 10 short questions about your business and it hands you three things: a production-grade system prompt for a WhatsApp AI agent, a ready-to-paste n8n workflow (SocialMate Trigger → AI Agent → tool nodes, with your own systems already wired in as agent tools), and a Claude Desktop / MCP config. No AI generates any of it — SocialMate assembles a hand-authored template from your answers, which is why it is free, instant, works offline and comes out identical every time. The AI is your own, running in your n8n or your Claude.
What you get
Not a prompt. A working agent.
Three artifacts come out of the last step — the brain, the plumbing, and the desktop path. Copy, paste, add credentials, go.
A ready-to-paste n8n workflow
Copy the JSON, paste it onto an n8n canvas: SocialMate Trigger → AI Agent → tool nodes, with the systems you named (Google Sheets, Shopify, HubSpot, Calendar…) pre-wired as agent tools. Your credentials are the only thing left to fill in.
A Claude Desktop / MCP config
The same agent, for the MCP path: a config block you drop into Claude Desktop, Cursor or any MCP client so it can drive WhatsApp as native tools.
The 10 steps
- Business. What you sell, who you sell to, your hours, your timezone and your website.
- Language. Answer in whatever language the customer wrote in, switching when they switch — on by default. Pick the languages you support, the one to fall back to, and the formality of each: German du/Sie and Japanese keigo are different systems, so one global setting would collapse the register in at least one of them.
- Agent. Name your agent and pick its role (support, sales, booking, orders, leads — or describe your own) and tone.
- Behaviour. How human it should feel: typing indicator, read receipts, reply delay, whether it splits long replies, and how it uses emoji. These are opinionated defaults you can change — not rules.
- Jobs. Tick the automations it should handle — from order confirmations and delivery updates to appointment booking, FAQ deflection and handing off to a human. Each one tells you exactly what it needs to work.
- Systems. Name the systems it should reach: Google Sheets, Airtable, Notion, Postgres, Shopify, WooCommerce, Stripe, HubSpot, Google Calendar, Google Drive, Slack — or anything else over HTTP. Each becomes a tool node in the generated n8n workflow, and the ones your chosen jobs actually need are suggested for you.
- Memory. Decide what it is allowed to remember: the conversation, the person, and what a photo or voice note said. Each one shows you the exact tool your AI gets and the SocialMate endpoint behind it. On Free the prompt honestly tells the agent it has none.
- Tools. Exactly what your agent can call. The jobs you picked already attached what they need — a booking job gets the poll read-back, a photo gets the media reader — so this step is for the extras: seeing what customers send, confirming a queued message landed, working with groups. Grouped into capability packs rather than a wall of switches, because an overlapping tool list is the top cause of an agent reaching for the wrong tool.
- Guardrails. AI disclosure (on by default), what it must escalate to a human, what is out of scope, and your out-of-hours message.
- Generate. Copy the seed system prompt, the n8n workflow JSON and the MCP config. Instantly, offline, with no model call in between — then edit the prompt and make it yours.
SocialMate does not generate this with AI — on purpose. Agent Studio is a template generator, not a model. It assembles a hand-authored, production-grade template from the answers you gave it. Nothing leaves your machine, nothing is billed, nothing is called — so it is instant, works with the Wi-Fi off, and gives you the same result on every run instead of a slightly different prompt each time. The AI is yours: it runs in your n8n or your Claude, and SocialMate is the WhatsApp hands and memory it uses. (AI generation by SocialMate itself is still on the roadmap — we will say so when it ships.)
21 automations
Tick the jobs it should do.
Each job you tick adds its rules to the prompt, attaches the tools it needs, and emits its own workflow. Jobs your tier can’t run stay visible, chipped — so you can see exactly what Pro would add.
Transactional — Meta’s own taxonomy
Seeded from the use-case enums in Meta’s WhatsApp Template Library (ORDER_CONFIRMATION, SHIPMENT_CONFIRMATION, PAYMENT_DUE_REMINDER, FEEDBACK_SURVEY…), so the naming of the transactional messages isn’t ours to argue about.
- Order confirmation
- Shipment confirmation & tracking
- Delivery update
- Order delay
- Ready for pickup
- Action needed on an order
- Order or transaction cancelled
- Return confirmation
- Payment reminder / overdue Pro
- Payment confirmation / receipt Pro
- Feedback survey (NPS / CSAT) Pro
Conversational & lifecycle
Meta’s library is Utility-only, so it stops here. This is our extension — and where most of the actual value lives.
- Answer questions from your knowledge
- Hand off to a human (with auto-resume)
- Order status (when they ask)
- Book an appointment
- Appointment reminders & rescheduling Pro
- Qualify and capture leads
- Cash-on-delivery confirmation Pro
- Abandoned cart recovery Pro
- Re-engagement / win-back Pro
- Tell customers who are waiting on you Pro
It writes your EU AI Act disclosure for you
Article 50 of the EU AI Act requires that people are informed they are interacting with an AI system — at the latest at the time of the first interaction (Article 50(5)) — and it becomes applicable on 2 August 2026. There is an exemption when it is “obvious”, but an agent deliberately built to sound human is exactly what removes it.
So Agent Studio writes the disclosure line into your system prompt, and it is on by default. You can switch it off — but even then the generated prompt forbids the agent from claiming to be a human, because lying about it is never the right call.
This is general information, not legal advice. Check your own obligations for your market and use case.
The generator is free. The memory is the upgrade.
Agent Studio itself never asks you for money. What changes with Pro is what the agent it built is actually allowed to do — and the prompt tells the truth about it either way. Generate on Free and the agent is instructed, in writing:
You have no memory of past messages. Each message you receive is all you know.
Never imply you remember an earlier conversation.That is not a nag — it is accurate. On Free the agent genuinely cannot read the archive, and an agent that thinks it can will confidently invent what a customer told it last week. Pro is what makes the line go away:
- Conversation memory — one call hands the agent the whole thread, role-mapped and token-windowed, so it never re-asks a returning customer something they already told you.
- Contact memory — when the agent learns a name, an email or a preference, it saves it against the contact and never asks again.
- Media memory — your model describes a photo or voice note once; SocialMate caches the description so the same file is never analysed twice.
- Scheduled & batch sends — the jobs that are reminders by nature — payment reminders, appointment reminders, cart recovery, win-back, telling customers who are waiting on you — need a queue to fire later. That is Pro.
SocialMate stores and serves what YOUR agent learned. It never generates it — no model of ours reads your chats.
Frequently asked questions
What is a WhatsApp AI agent prompt generator?
A tool that turns a few answers about your business into the system prompt that makes an AI agent behave correctly on WhatsApp. SocialMate’s is called Agent Studio, it is built into the app (sidebar → AI Agent), and it also emits the plumbing: a ready-to-paste n8n workflow and a Claude Desktop / MCP config, so you get a working agent rather than a block of text.
Does SocialMate use AI to write the prompt?
No — and that is deliberate. Agent Studio assembles a hand-authored, production-grade template from your answers. Nothing is sent anywhere and no model is called, which is exactly why it is free on every tier, instant, works offline, and produces the identical result every time instead of a slightly different prompt on every run. The AI is yours: it runs in your n8n or your Claude. (AI generation by SocialMate itself is still on the roadmap, not shipped.)
Is Agent Studio free?
Yes — it is free on every tier, including the Free tier, with no account and no API key. Download SocialMate, open the AI Agent page, answer the 10 steps, copy your prompt. What changes on Pro is not the generator — it is what the generated agent is allowed to do.
What exactly do I get at the end?
Three artifacts. (1) A system prompt that names the exact tools your agent has, tells it how to back off when a send is rate-limited, states what it must never claim, and defines its escalation path to a human. (2) An n8n workflow as JSON — paste it onto the canvas and you get SocialMate Trigger → AI Agent → tool nodes, with the systems you named already wired in as agent tools; your credentials are the only TODO. (3) A Claude Desktop / MCP config for the MCP path.
Do I need to know n8n to use it?
Not really. Copy the generated JSON, open an n8n canvas and paste — the nodes, the wiring and the agent’s tool list arrive already connected. You then fill in credentials for the services you picked (your Google account, your Shopify store, your OpenAI or Anthropic key) and switch it on.
Does my agent have to tell people it is an AI?
Under the EU AI Act, yes — Article 50 requires that people are informed they are interacting with an AI system, at the latest at the time of the first interaction (Art. 50(5)), and it becomes applicable on 2 August 2026. Agent Studio writes that disclosure line into the prompt for you and it is on by default. You can switch it off, but the generated prompt still forbids the agent from claiming to be human.
Which automations does it cover?
21, and you tick the ones you want. The transactional set is seeded from Meta’s own WhatsApp Template Library use-case taxonomy — order confirmation, shipment confirmation, delivery update, order delay, ready for pickup, action needed, cancellation, return confirmation, payment reminder, payment confirmation, feedback survey. Meta’s library stops at Utility messages, so SocialMate adds the conversational and lifecycle work: FAQ deflection from your own knowledge, human handoff with auto-resume, order status, appointment booking and reminders, lead capture, cash-on-delivery confirmation, abandoned-cart recovery, win-back and telling customers who are waiting on you.
What changes on Pro?
Not the generator — the agent. On Free the generated prompt literally instructs the agent: you have no memory of past messages, never imply you remember — because on Free it genuinely cannot read the archive. Pro unlocks conversation memory (one call hands the agent the whole thread), contact memory (it saves the name or preference it just learned, and never re-asks), media memory (your model describes a photo once and the description is cached), plus scheduled and batch sends. Jobs that need those stay visible on Free with a PRO chip so you can see exactly what you would gain.
Will running an AI agent get my number banned?
Automating WhatsApp can violate its Terms of Service, and no tool is 100% ban-proof. Every send an agent makes goes through SocialMate’s anti-ban engine — rate limits, jitter, warming, live risk scoring — so the agent cannot burn the number with a retry loop, and the generated prompt teaches it to back off rather than hammer a blocked send. Use it with people who expect to hear from you.
Your agent is 10 answers away.
Agent Studio ships in every SocialMate install, on every tier. Download it, open the AI Agent page, and copy your prompt.