ServiceChat — Customer Conversations, Unified
ServiceChat
Ticket Center
ServiceChat

One inbox for web, social and AI conversations

Bring your website Widget, social DMs, AI memory, human handoff and Bot API into a single inbox — so your team can actually keep up with real customer conversations.

Support inbox 28 open
Linda · LINE
typing
Does the Pro plan integrate with our CRM?
2m
Kenji · Telegram
I have a question about shipping
9m
Omar · WhatsApp
Has my order shipped yet?
14m
Sophie · Web Widget
Do you offer API access on the Pro plan?
23m
AIAI drafted 3 replies — waiting for an agent to send.
LIVE IN FIVE MINUTES

Paste one snippet. Start taking conversations today.

Web, LINE, Telegram, WhatsApp and Messenger into one inbox, where your team answers them together. The free tier takes no card.

Create a free account
Shared inbox
One workspace for the whole team
1 GB media
Images, video and voice — all in chat
Any language
Auto-translation and AI replies
Works everywhere
Desktop, tablet and mobile
PRODUCT

Put live chat on your site in 5 minutes

Start on the free plan. Verify your email, paste a Widget snippet into your site, and you'll receive your first test message. Add social channels and AI replies whenever you're ready.

01

Sign up & verify

Enter your company details, confirm via email and pick a password.

02

Embed the Widget

Paste one script tag into your site — the chat window shows up immediately.

03

Start replying

Messages land in your inbox the moment a visitor sends them. Add social channels and AI later.

SHARED INBOX

One workspace your whole team works from

Open, in progress, closed, unread, customer profile — all in the same view. Switch between conversations without losing context.

  • Full conversation history
    Every past message, order and tag for each customer in one place.
  • Assign and hand off
    Route conversations to the right teammate, or take over from the AI.
  • Statuses & tags
    Track progress with statuses like In progress, Waiting and Resolved.
  • Real-time alerts
    New messages and unread counts pushed to desktop and mobile.
L
Linda Chen
LINE · 3rd conversation
In progress
Does the Pro plan integrate with our CRM?
Yes — Pro includes Bot API for connecting to your existing CRM.
Got it. Any docs on setting up the webhook?
Agent is typing
yourshop.com
1
CUSTOMIZABLE WIDGET

A chat window that looks like part of your brand

Customise colour, greeting, avatar and placement. From a quick palette tweak to full white-label, the Widget feels native to your site.

Brand colour
Pick from presets or paste a HEX value.
Launcher style
Five built-in styles, or roll your own via launcher API.
Avatar & position
Swap in your own agent avatar or brand logo.
Display rules
Show on specific pages or only during business hours.
TICKET CENTER

When chat isn't enough, open a ticket and see it through

Create a ticket in one click from any conversation, or put a form on your own site and every submission becomes a ticket. Fields, statuses, notifications and system sync all live in one Ticket Center.

  • Web forms straight to tickets
    Put a form on your site or help center; each submission opens a ticket, with email verification to confirm the sender.
  • One-click tickets from chat
    Open a ticket right inside the conversation — customer identity and the full transcript attach automatically.
  • Your forms, your workflow
    Every ticket type gets its own fields, statuses and closing rules, defined around your business.
  • Two-way sync with your systems
    Ticket changes push to JIRA or your CRM in real time, and their status updates and comments write back.
Ticket Center
+ New ticket
TK-24081
Refund status inquiry
In progress
TK-24079
Can't sign in
Waiting on customer
TK-24075
Invoice request
Closed
A form on your site
Name
Email
Describe the issue
Submit
Auto-creates a ticket
HOW THE AI WORKS

AI support is not “forward the message to a model”

Every customer message walks a full pipeline: decide whether the AI should even speak, bring memory and knowledge along, verify before answering — and hand off properly to a human when it cannot answer, instead of guessing.

1
Message arrives
Website, LINE, Telegram, WhatsApp, Messenger — every channel and every language enters the same pipeline.
2
Gatekeeping
Has a human taken over? Is the customer still typing? Inside service hours? The AI first decides whether to speak at all.
3
Context assembly
Cross-conversation customer memory, cross-lingual knowledge retrieval and your custom skills are loaded into this one reply.
4
Tools verify
Look up the order, call your APIs, actually test the website — get the facts before earning the right to answer.
5
Automation decides
Rules plus judgement decide the next step: reply, open a form, or hand off to a human.
Three outcomes
AI replies
Answers from knowledge and verified results, in the customer’s language.
Guided form
When data is needed, open a form instead of interrogating back and forth.
Human handoff
Category, reason, summary and collected data travel with it — the agent starts warm.
Every turn is written back into customer memory — next conversation, the AI picks up where it left off, and the customer never repeats themselves.

The architecture behind that flow

A bounded agent: the core loop is wrapped in code-enforced gates, resource layers and per-item validation — every layer with an explicit mandate.

Customer message arrives
Website widget, LINE, Telegram, WhatsApp, Messenger — any channel, any language.
Agent console
Humans first: the moment an agent takes over or is typing, the AI steps back.
↓
Decision layer — code-enforced gates
Humans first · service hours · reply rate caps · at-most-once execution per trigger
Enforced in code. The model has no authority to cross this layer.
↓
AI Agent core — plan → act → observe loop
Six specialized model roles: action planner · tool selector · result summarizer · memory curator · decision router · query translator — each output validated independently.
Step caps · fingerprint dedup on identical calls · every cited code verified
↕ read / write
↕ invoke
↕ delegate
Context layer
Cross-conversation customer memory · cross-lingual knowledge retrieval — Chinese docs answer Thai questions.
Tool layer
Order APIs · code sandbox · real browser tests — facts first, then answers.
Sub-agent delegation
Engineering issues go to a read-only investigation agent inside the merchant’s own environment, which reports back.

Guaranteed by architecture, not by prompts

“Say you don’t know” written in a prompt is a wish. ServiceChat writes every rule into the code layer — design is the part where the model doesn’t get a vote.

What answers are based on
✕Throwing messages at a model
The model improvises from training memory. The prompt begs it to “say you don’t know” — but a model doesn’t know what it doesn’t know.
✓How ServiceChat does it
Cross-lingual knowledge retrieval and tool verification run before any answer. Every form code or handoff rule the model cites is checked in code; anything that doesn’t match degrades to a human handoff. Fabrication is intercepted before sending — not left to the model’s conscience.
How decisions are made
✕Throwing messages at a model
One giant prompt handles understanding, lookup, decision and reply — when it fails, nobody can say where.
✓How ServiceChat does it
Six specialized model roles, each owning one step: planning, tool selection, result summarizing, memory curation, decision routing, query translation. Each output is validated independently, so one failure never drags down the whole line.
Ability to act
✕Throwing messages at a model
The model “says” it checked — nothing actually happened.
✓How ServiceChat does it
A plan → act → observe agent loop: it really calls your order API, runs code in a sandbox, opens a browser to test your site, and feeds results into the next step. Steps are capped and identical calls are fingerprint-deduplicated — no spinning in circles, no double charges.
Memory
✕Throwing messages at a model
Every conversation starts from zero; the customer retells the story.
✓How ServiceChat does it
Every turn writes back into cross-conversation customer memory — handoffs and closures included. Next conversation, the AI continues from where it left off.
When it’s stuck
✕Throwing messages at a model
Apologize, or fabricate something.
✓How ServiceChat does it
Engineering issues are delegated to a read-only investigation agent inside the merchant’s own environment, which reports back for the support AI to compose the reply. What it can’t answer goes to a human — with category, reason, summary and collected data attached.
Who guards the boundaries
✕Throwing messages at a model
Hope the model restrains itself.
✓How ServiceChat does it
Humans first (the AI steps back the moment an agent takes over), reply rate caps, at-most-once execution per trigger, service-hour gates — all enforced in code, outside the model’s decision space.

Why this design actually helps your customers

Customers don’t want “a response” — they want to be handled. These choices are the difference.

No pretending to know
Every answer passes retrieval and tool verification first. When nothing is found, the AI says so and hands off — it never invents an answer to brush a customer off.
Humans always win
When an agent has taken over or is typing, the AI steps back automatically. Your customers are never fenced off from a human by a bot.
It remembers every customer
Last time’s question, the half-finished case — the AI continues from there. The dreaded “please describe your issue again” never happens.
No language boundaries
Knowledge base in Chinese, customer asking in Thai or Russian — retrieval still works. No fixed language list; any pairing holds.
It acts, not just talks
Order status, API diagnostics, website checks — the AI actually looks things up and tests them, answering with results rather than reciting documents.
Handoffs never drop the ball
The moment a human takes over, the category, reason, summary and collected data arrive with it. Agents skip the archaeology; customers skip the retelling.
BOT API

Need automation? Hand the API to your developers

Want to plug in your own AI, CRM, order system or ERP? ServiceChat ships a full Bot API. Your devs write a webhook and conversations flow back into the systems you already use.

  • →REST API and Webhooks for real-time events
  • →Wire up your own AI agent or LLM provider
  • →Two-way sync for customer and order data
  • →Complete docs, samples and a sandbox env
POST /api/public/v1/rooms/{roomId}/messages/customer-text
curl -X POST https://servicechat.ewin888.com/api/public/v1/rooms/$ROOM_ID/messages/customer-text \n  -H "X-ServiceChat-ApiKey: $SC_API_KEY" \n  -H "Content-Type: application/json" \n  -d '{
    "text": "Hi — your order has shipped."
  }'

// // Response
{
  "success": true,
  "data": {
    "messageId": "9f2c…",
    "createdAt": "2026-08-31T09:15:22Z"
  }
}
WHO IT'S FOR

Built for teams stitching together scattered support channels

Got a website, social accounts, and messages piling up in five different apps? These are the teams that get the most out of it.

E-commerce & retail

Order questions, returns and shipping updates — all in one queue.

Service businesses

Bookings, enquiries, member support. AI handles the basics, humans close the case.

Cross-border brands

Customers write in any language; the reply comes back in the same one.

Games & live ops

Top-ups, account issues, event questions — answered 24/7.

Agencies & integrators

Set up support for your clients, or run multiple brands from a single console.

SaaS & tools

Tier issues by severity; engineers can plug in directly via API.

WHY SERVICECHAT

More than a chat widget — it's where support actually runs

Live in 5 minutes

No engineering ticket required — your marketer can paste the script in.

Human + AI together

AI takes the routine questions; humans step in for the tricky cases.

Context never gets lost

Customers change channels, agents change shifts — history travels with them.

Wire it back to your stack

Two-way sync with your CRM, ERP, orders and ticketing via Bot API.

Quality you can measure

First-reply time, resolution time and CSAT trends, right in the console.

Clean access boundaries

Agents see only their own conversations; managers see the whole picture.

SECURITY

Clean access boundaries by design

Agent accounts, API Keys, Widget Keys and conversation data are managed separately. Teammates touch only the conversations they need — nothing more.

Tenant isolation
Each brand's conversations and knowledge files are stored on their own, never visible to another merchant.
Role-based access
Three tiers: admin, manager, agent.
Encryption in transit
TLS 1.3 across every connection.
Audit logs
Every console action is logged and queryable.
Admin Manager Agent A Agent B
CUSTOM DOMAIN

Your chat entrance, on your own domain

From Starter up, register your own domains for the chat entrance, with a limit per plan. The page carries one link and no address of ours; if a domain is blocked, switch it in the console and the page stays as it is.

Your domain
One CNAME record; the certificate is issued for you.
No address in the source
The link is encoded, so the page never shows where the chat lives.
Blocked? Swap it
Add or replace a domain in the console; the customer's page never changes.
PRICING

Four plans — pick the one that fits

From solo merchants to multi-brand enterprises. Bring your own provider key for AI features.

Free

For solo merchants validating their support flow.

US$0
  • 1 agent
  • Website Widget
  • Text & image messaging
  • Basic welcome message
  • Customer profile card
  • ServiceChat branding
Get started

Starter

Small merchants going live with web support.

US$29/ month
  • 3 agents
  • Website Widget
  • Images, video, voice and files
  • Resumable large file uploads
  • Messenger bots
  • AI translation
  • Self-serve menus
  • 1 Bot API Key
  • 1 custom domain
Get started
Most popular

Pro

Teams with real support workflows and AI automation.

US$89/ month
  • 10 agents
  • AI agent & knowledge base
  • AI memory & summaries
  • Cross-conversation customer memory
  • Pinned files
  • Agent performance & CRM data
  • Custom launcher & unread events
  • 3 custom domains
Get started

Enterprise

Multi-brand, large teams and dedicated support.

Custom
  • Custom agent seats
  • Advanced translation settings
  • Custom AI knowledge base
  • Multi-layer self-service flows
  • Unlimited custom domains
  • Audit & security controls
  • SLA & integration support
Get started
AI translation and AI replies require your own provider key (e.g. OpenAI / Anthropic). Provider costs are billed separately.

A few fields — and
your tenant is ready

Just a few fields. We'll email you a verification code; once you've set a password you can sign in, install the Widget and create an API Key.

1
Fill in details, get the code
Verification code goes to the email you provide.
2
Set a password, create the tenant
The admin account signs in directly.
3
Install Widget & create API Key
Paste the script onto your site and generate a developer API Key.
FREE TO START

By submitting you agree to our Terms & Privacy Policy。

Service created

Sign in with the registered email, then create API keys in the API Center.

Go to Console Login

Frequently asked

Can I install it without a developer?+
Yes. Copy a script tag into your site's HTML — that's it. Shopify, WordPress and Wix are all supported, with step-by-step guides in the console.
Does it integrate with LINE, WhatsApp and Telegram?+
Starter and above support website Widget, Telegram, LINE, WhatsApp, Messenger and other major channels. We're adding more.
How does the AI agent work? What do I need?+
You provide FAQs, product info or documents — the AI replies based on them. You bring your own OpenAI or Anthropic API key (provider fees billed separately). When the AI isn't sure, it hands off to a human.
How is the AI different from a normal FAQ bot?+
ServiceChat combines conversation context, long-term summaries and cross-conversation customer memory — so it isn't just reciting a manual.
Can the AI agent handle group chats too?+
Yes. Mention the agent in a group thread to trigger the AI; the system stitches together the surrounding context to avoid misreads from messages split across multiple lines.
Can multiple people use one tenant at the same time?+
Yes. Paid plans are sold by seat — each agent signs in separately with their own role and permissions. The team can handle different conversations in parallel and hand off as needed.
How long is message data stored? Can I export it?+
Retention depends on plan — 12 months on Starter and Pro, extended (or dedicated) on Enterprise. Conversations can be exported as CSV or JSON via the console or API.
Will large video uploads have to start over if interrupted?+
No. The Widget supports chunked, resumable uploads.
Will my conversations be used to train AI?+
No. Your data belongs to your account. The AI features run through the provider key you bring, and training behaviour follows that provider's policy (OpenAI and Anthropic's business APIs default to no training).
Which languages do you support? How good is the translation?+
Seven built-in languages: Traditional Chinese, Simplified Chinese, Japanese, Korean, Thai, Vietnamese and English. Translation runs on LLMs — context and terminology handling is better than traditional MT.
What happens when we exceed the agent limit on our plan?+
Adding an agent prompts you that your plan is full — you can buy extra seats or move up a tier.
What happens when we change plans mid-cycle? Can we downgrade?+
Upgrades take effect immediately and all data is preserved. Downgrades take effect at the end of the cycle; old data stays viewable and exportable, but some advanced features turn off.
How is this different from Zendesk or Intercom?+
ServiceChat is focused on web + social + AI + human in one place — lighter UI, faster onboarding, pricing closer to the Asian market. Need advanced ticketing or CRM? Our Bot API connects to your existing stack.
How do developers integrate? Is there an SDK?+
We provide a full REST API and Webhook reference, plus JavaScript / Python / PHP samples. Test in the sandbox, then go live. Docs are public; API Keys are one click in the console.
What about custom integration?+
Enterprise plans include a dedicated CSM and technical support. We can help scope CRM, ERP and POS integrations — or move you to a dedicated/private deployment.