# ChatFlowGate for n8n

> ChatFlowGate is a security gateway and white-label chat frontend for n8n Chat
> workflows. Visitors talk to ChatFlowGate, which authenticates, rate limits, and
> meters every message before forwarding it to the customer's n8n webhook.

Site: https://www.chatflowgate.com/

## What it is

n8n builds the assistant. ChatFlowGate is the layer between the public and that
workflow, so a chat trigger can be exposed to real users without publishing the
webhook URL, absorbing abuse, or building auth, rate limiting, and analytics by
hand. It is a hosted product with a free tier; the widget embeds with one
script tag.

## Technical architecture

- Hidden webhook: the n8n URL is stored server-side and is never sent to the
  browser. The page only ever calls the ChatFlowGate chat endpoint for a bot id.
- Caller identity, most trusted first: an organization-scoped API key sent as
  X-API-Key, a signed-in member of the owning organization, or a visitor token.
- Visitor tokens: HMAC-SHA256, bound to a single bot id, expiring in 24 hours,
  issued only after the parent page origin is checked against that bot's domain
  allowlist. In lead capture mode a token is issued only after the contact form
  is submitted, so the token itself proves the form was completed.
- Rate limiting: token buckets per session and per IP, configured per bot,
  answering HTTP 429 with a Retry-After header. Shared across instances through
  Redis when configured, per-instance otherwise. The client IP is read a fixed
  number of proxy hops from the right of X-Forwarded-For, so a forged header
  cannot mint a fresh quota.
- SSRF protection: webhook targets are DNS-resolved and rejected when they point
  at loopback, private, link-local, or cloud metadata addresses. Redirects are
  not followed. A self-hosted n8n must therefore be reachable on the public
  internet.
- Abuse controls: organization-scoped IP bans checked before any other work,
  per-bot country allow and block lists, and hidden spam-trap fields on the
  widget forms that are checked again server-side.
- Metering: one message spends the monthly plan allowance first, then purchased
  top-up credits, in a single atomic statement. HTTP 402 when both are empty.
- Data retention: message content is never stored. One row per chat session
  holds IP, country, region, city, parsed browser, OS, device, and a message
  counter, plus submitted contact details in lead capture mode.
- Multi-tenancy: bots, sessions, API keys, IP bans, credits, and members belong
  to an organization, and every read and write is scoped to it.
- Streaming: replies from n8n are parsed line by line (NDJSON token chunks, SSE
  data frames, or a single JSON body) and streamed to the widget as text deltas.
- Management API: an MCP server at /api/mcp exposes bot listing, creation,
  updates, and analytics to AI clients, authenticated with an API key.

## Primary use cases

1. Exposing an n8n chatbot on a public website without publishing the webhook.
2. Capping abuse and runaway model spend on a public chat workflow.
3. White-labelling and reselling n8n chatbots to clients from one dashboard.
4. Capturing leads before a conversation starts and forwarding them to n8n.

## Plans

Free $0/month: 1 bot, 1 seat, 500 messages, ChatFlowGate badge shown.
Lite $19/month: 3 bots, 1 seat, 5,000 messages, badge removable.
Pro $29/month: 10 bots, 3 seats, 10,000 messages, badge removable.
Max $99/month: 30 bots, 6 seats, 50,000 messages, badge removable.

Top-up credit packs cover usage past the monthly allowance and do not expire.

## Limits and honest caveats

- ChatFlowGate does not store or replay conversation history. A transcript archive
  is not a feature.
- A self-hosted n8n on localhost or a private network address cannot be reached
  by design; the SSRF guard blocks it.
- ChatFlowGate holds no formal security certification such as SOC 2 or ISO 27001.