SocialMate

v2.7.1

Last updated Jul 14, 2026

n8n community node for SocialMate — automate WhatsApp messaging through your self-hosted SocialMate desktop server (send, schedule and queue personalized messages, manage chats, contacts, groups, media, and react to incoming messages).

6 Weekly Downloads

84 Monthly Downloads

View on NPM GitHub Repository

Included Nodes

SocialMate

SocialMate Trigger

Description

n8n-nodes-socialmate

node version

Official n8n community node for SocialMate — the self-hosted WhatsApp automation server. Run it as a desktop app (Windows/macOS/Linux) or headless on your own VPS with Docker; this node talks to either the same way. On the desktop it sends from your own residential IP (same network as your phone); on a VPS (a datacenter IP) you can route an account through your own residential/mobile proxy — set it in the app//admin or via PUT /v1/accounts/:id/proxy (Pro).

Automate WhatsApp from n8n: send and schedule personalized messages, queue and batch-import to opted-in contacts (an opt-in Pro feature), manage chats, contacts, groups and media, and react to incoming messages and lifecycle events — all through your own SocialMate server.

Docs: https://socialmate.app/docs/n8n
Changelog: CHANGELOG.md

Installation

In n8n: Settings → Community Nodes → Install and enter n8n-nodes-socialmate.

Or self-hosted via npm:

npm install n8n-nodes-socialmate

Connecting

1. In the SocialMate app: API & Integrations → n8n → New connection. Pick the account
scope, and SocialMate mints an API key (and a self-registering webhook) for it. Use an
admin-scope key if you want the Trigger to register its webhook automatically.
2. In n8n, create a SocialMate API credential:
– Server URL — http://127.0.0.1:3456 when n8n runs on the same machine as the
app, or your named-tunnel hostname (Pro) for a remote n8n. No trailing slash.
– API Key — the key from step 1.
3. Click Test — you should see your server version and license tier.

> Quick tunnels are not supported. A free quick tunnel gets a new
> *.trycloudflare.com URL on every restart, which breaks a saved credential. Run n8n on
> the same machine as the app (localhost) or use a stable named tunnel (Pro).

Account scope (v2.1+)

Every SocialMate API key carries an account scope, chosen when you create the
connection in the app:

There is no “Default Account ID” — the key already knows which accounts it may use:
GET /v1/accounts is filtered to that set server-side, so the node’s Account dropdown
only ever shows accounts you can actually use, and auto-resolves when the key is bound to a
single account. The SocialMate Trigger registers its webhook with the same key, so the
server scopes deliveries automatically (an optional account filter exists to further narrow
an “All accounts” key).

> Upgrading from v2.0: an old credential’s Default Account ID is still honored as a
> fallback, but you can clear it — the scope-aware Account picker replaces it.

Nodes

SocialMate (action) — Message, Chat, Contact, Group, Media, Queue, Account, Sync, Webhook, API Key and System operations.
SocialMate Trigger — fires your workflow on SocialMate events. It self-registers its webhook with the server and verifies the HMAC signature on every delivery.

Operations — Free vs Pro

Operations marked Pro require a SocialMate Pro license; on Free they return a clear
402 naming the missing feature.

Every send still passes through SocialMate’s anti-ban engine (rate limits, jitter, warming,
risk checks), so a runaway workflow can’t burn your number.

Build an AI agent with full chat memory — `Message → Get AI Context`

The Trigger only hands your workflow the single new message. To give an AI agent the
whole conversation, drop Message → Get AI Context between the Trigger and your
AI Agent. It reads the chat’s history from the SocialMate DB and returns it AI-ready:

transcript — a role-mapped string to paste straight into the agent’s system message:

Conversation so far:n{{ $json.transcript }}.

messages — a structured [{ role: 'user'|'assistant', name, content, ts }] array for

chat-model / memory nodes.

meta — { totalMessages, returnedMessages, truncated, tokenEstimate, … }.

It role-maps automatically (the contact = user, your account = assistant), windows the
history to a token budget (maxTokens, default 4000) and message cap (maxMessages,
default 50), supports order (oldest | newest), format (both | messages |
transcript), includeTimestamps, and beforeTs (window to messages before a Unix-ms time),
and labels media ([image], [voice],
…). (Requires SocialMate Pro — reading history.) See
examples/real-estate-deepseek-agent.json.

Use SocialMate as an AI Agent tool

Every SocialMate operation is exposed as an AI Agent tool (usableAsTool), so an n8n **AI
Agent** can call them on its own — _”send this reply”, “look up the contact”, “fetch the last 20
messages”, “queue a reminder”_ — choosing the operation from its description.

Enable it on your n8n instance (required for any community node used as a tool):

N8NCOMMUNITYPACKAGESALLOWTOOL_USAGE=true

Then add SocialMate as a tool on an AI Agent node and pick the operation. Pair it with
Message → Get AI Context (the conversation-memory tool) so the agent can read a whole thread
before replying — a combination no official / Business-API node offers.

> Some n8n versions have an open bug where certain community-node tools hand the agent an empty
> observation (n8n#26202). If you hit it, upgrade n8n,
> call the operation from a normal (non-agent) node, or use MCP (below), which bypasses it.

Prefer MCP (Claude Desktop / Cursor / any client)? SocialMate ships a native **Model Context
Protocol server, socialmate-mcp, exposing 44 WhatsApp tools (see API & Integrations → MCP** in
the app for the copy-paste config), plus an n8n MCP Server Trigger recipe. One thing to know: MCP is
request/response with no inbound push, so to auto-react to incoming messages you either poll the
whatsappfetchnew_messages tool or drive the loop from the SocialMate Trigger below. Full
walkthrough + the $fromAI pattern + 11 use cases in
docs/AI-AGENT-TOOL-GUIDE.md. Importable examples:
examples/ai-agent-tool.json and
examples/mcp-server-trigger.json.

Make it behave like a human

A drop-in System Message that teaches an agent the human reply cadence — *mark read →
recall the thread → show typing → react or reply (threaded)* — plus the full tool inventory,
the tier and anti-ban error contract (402 / blocked / queueable:false /
signalratelimit), what the agent genuinely cannot do (no edit/delete/forward, it can’t
see the contact typing, buttons are deprecated → send a poll), and the consent + honesty rules:

docs/AI-AGENT-SYSTEM-PROMPT.md

examples/ai-agent-tool.json ships with it pre-loaded and the tools wired
(Mark Read · Show Typing · React · Send Poll · Get Conversation Memory · Check My Limits).
Reactions, read receipts and the typing indicator are free on every tier and consume no
send budget — an agent can be human without spending its message allowance.

Trigger events

The SocialMate Trigger covers all 35 events. 9 are available on Free —
message.received, message.sent, account.connected, account.disconnected,
tunnel.url_changed, tunnel.stopped, license.activated, license.deactivated,
license.tier_changed; the other 26 (incl. tunnel.started, the delivery receipts
message.delivered / message.read, the conversational events message.reaction /
poll.vote / group.participantsupdated, the Agent Memory media.contextupdated event,
and the High-Volume Mode account.dangermode* events) require Pro and are labelled (Pro)
in the picker:

Messaging: message.received, message.sent, message.delivered (Pro), message.read (Pro), message.reaction (Pro), poll.vote (Pro)
Groups: group.participants_updated (Pro) — join, leave, promote, demote
Accounts: account.connected, account.disconnected, account.banned, contacts.updated, account.dangermodeenabled, account.dangermodedisabled
Tunnel: tunnel.started, tunnel.url_changed, tunnel.stopped
Sync: sync.started, sync.completed, sync.failed
Media: media.discovered, media.downloaded, media.failed, media.deleted, media.context_updated
Smart queue: queue.item.enqueued, queue.item.processing, queue.item.sent, queue.item.failed, queue.item.cancelled, queue.batch.created, queue.batch.completed, queue.batch.cancelled
License: license.activated, license.deactivated, license.tier_changed

> Delivery receipts. message.sent only means “handed to WhatsApp”. message.delivered and
> message.read close the loop — but a contact with read receipts turned off never produces
> message.read, so its absence is not proof the message went unread. Correlate on messageId.

Every delivery is a JSON envelope: { "version", "event", "timestamp", "tunnelUrl", "data": { … } }. A
message.received example:

{
  "version": 1,
  "event": "message.received",
  "timestamp": "2026-06-03T12:00:00.000Z",
  "tunnelUrl": "https://your-name.example.com",
  "data": {
    "accountId": "acc_123",
    "messageId": "ABCD1234",
    "chatId": "15551234567",
    "senderId": "15551234567",
    "body": "Is my order shipped?",
    "type": "text",
    "timestamp": 1750000000000,
    "fromMe": false
  }
}

In a workflow, read fields as {{ $json.data.body }}, {{ $json.data.chatId }}, etc.

Request & response examples

The action node builds these requests for you; the JSON below shows the underlying SocialMate
API shapes (handy when mapping fields). Responses are wrapped in the { "data": … } envelope.

Message → Send Text — when to use: the default outbound path; works on every tier.

// request
{ "chatId": "15551234567", "text": "Hello from n8n 👋" }
// 200
{ "data": { "sent": true, "messageId": "msg_abc", "chatId": "[email protected]", "status": "sent" } }

Message → Send Media (Pro) — prefer a url so a rate-limited send can auto-queue (202):

{ "chatId": "15551234567", "media": { "type": "image", "url": "https://example.com/p.jpg", "caption": "Invoice" } }

Message → Get AI Context (Pro) — when to use: give an AI agent the whole thread in one call.
Note tokenEstimate lives inside meta, not at the top level:

{
  "data": {
    "account": { "id": "acct_1", "name": "Sales", "phone": "15551230000" },
    "chat": { "id": "15551234567", "name": "Jane" },
    "messages": [ { "role": "user", "name": "Jane", "content": "Hi", "ts": 1750000000000, "type": "text" } ],
    "transcript": "Jane: Hi",
    "meta": {
      "totalMessages": 1, "returnedMessages": 1, "truncated": false,
      "oldestReturnedTs": 1750000000000, "newestReturnedTs": 1750000000000,
      "tokenEstimate": 8,
      "window": { "maxMessages": 50, "maxTokens": 4000 }
    }
  }
}

Queue → Queue a Batch (Pro; opt-in, off by default) — when several people who are **already
waiting on you need the same news (an order delay, a new pickup time), queue one individual,
personalised message each** and let the anti-ban engine pace them — up to 5000 rows. Only for people
who contacted you or explicitly opted in; never a list you bought, scraped or guessed. This is not a
broadcast: identical text to many contacts is blocked by the duplicate guard, so personalise every
row.

{
  "template": "Hi {{name}}, your order {{order}} shipped 🚚",
  "batchName": "June shipping notices",
  "rows": [ { "chatId": "15551234567", "displayName": "Jane", "fields": { "name": "Jane", "order": "A-1001" } } ]
}

Account → Get Anti-Ban Status — when to use: check headroom before a burst; if paused or a
counter is near its max, back off (sending anyway just auto-queues on Pro or returns 429 on Free).

API-key scopes

Keys carry one or more scopes; an operation that needs more than the key has returns 403:

read — GET operations (accounts, chats, contacts, groups, media, status…).
send — sending messages, the signal lane (react / mark read / typing), enqueuing to the

smart queue, and forcing a media download. Also the Agent Memory writes (Contact → Update,
Media → Set Context) — saving what your agent learned shouldn’t require account-admin rights.

admin — managing webhooks and API keys (the Trigger needs this to self-register),

triggering a Sync, and Set / Clear Proxy. The rule of thumb: read means *your WhatsApp
data; anything that changes how the server itself is wired* is admin.

Errors

| Status | Meaning |
|—|—|
| 401 | Missing or invalid API key. |
| 402 | A Pro feature on a Free license — the response names the required feature. |
| 403 | The key lacks the required scope (read / send / admin). |
| 429 | Rate-limited. A transient per-key limit is retried automatically (up to 3×, capped backoff); an anti-ban send block is surfaced as data instead — see Send outcomes. |

Send outcomes

A Send Text / Send Media call resolves one of these ways. The first three are returned as
data you can branch on (no error thrown), so one workflow can handle every case:

Sent — delivered immediately ({ sent: true, messageId, … }).
Queued (Pro) — anti-ban deferred it, so SocialMate auto-queued it and the worker

retries when the number is eligible ({ queued: true, itemId, … }).

Blocked (typically Free) — anti-ban refused it and there’s no auto-queue, so the node

returns { blocked: true, reason, retryAfterMs, hint, upgrade }. Branch on blocked to back
off, notify, or (on Pro) route to Queue → Enqueue. The node returns immediately — it
never sleeps on the block’s Retry-After (which can be hours during quiet hours).

409 — the account isn’t connected; link it in the app first.

> The legacy POST /v1/accounts/:id/messages/media route is deprecated (no auto-queue,
> sends a Sunset header) and is intentionally not exposed as an operation — use **Send
> Media** (the unified endpoint) instead.

License tiers & limits

Free covers text send, reads (chats/contacts/accounts/media), live webhooks, and
webhook/API-key management — one account. Pro adds media send, group management,
scheduling and opt-in batch sending, message history/search, sync, the smart queue, unlimited accounts
and a stable named tunnel.

Free is capped at 200 sends/day; Pro starts at 500/day per account and scales to
up to 5,000/day in High-Volume Mode (after 72h warming).

Examples

Importable workflows live in examples/:

ai-agent-tool.json — SocialMate as an AI Agent tool: an AI Agent that marks read, recalls the thread, shows a typing indicator, then reacts or replies (threaded). Ships with the human-cadence system prompt pre-loaded. See docs/AI-AGENT-SYSTEM-PROMPT.md and docs/AI-AGENT-TOOL-GUIDE.md.
mcp-server-trigger.json — WhatsApp MCP server built in n8n: expose SocialMate to Claude Desktop / Cursor via the MCP Server Trigger.
real-estate-deepseek-agent.json — a WhatsApp AI concierge (Trigger → Get AI Context → DeepSeek → reply).
auto-reply-trigger.json — a minimal Trigger → Send Text auto-reply round-trip.
scheduled-bulk-send.json — schedule a daily batch of queued, personalised reminders to opted-in contacts via the smart queue (Pro). One individual message per person, paced by the anti-ban engine — not a broadcast.
keyword-router.json — Trigger → IF on the message text → route to different replies (extend with more branches).
drip-onboarding-sequence.json — Trigger → three scheduled Queue Enqueues that drip a welcome sequence over a few days (Pro).
order-confirmation-webhook.json — your store’s order webhook → Send Text order confirmation.
new-lead-alert.json — Trigger → forward a short summary of each inbound message to your team group.
vision-memory-loop.json — Agent Memory (Pro): media.discovered → skip already-described items → Download File → your vision/transcription model → Set Media Context. Each photo/voice note is described once and then comes back as [image: …] inside Get AI Context — no re-analysis, fewer tokens.

Development

npm install
npm run build      # tsc + copy icons to dist
npm run lint       # eslint-plugin-n8n-nodes-base

License

MIT