ChatterMate MCP server
Commit Nov 14, 2026An open-source AI customer support platform that automates queries via omnichannel chat and connects to external tools using MCP.
Also known as chattermate/chattermate.chat, chattermate-chat
- Stars
- 168
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 14, 2026
- License
- Apache-2.0
What is ChatterMate MCP?
ChatterMate is an open-source AI customer support platform that automates queries via omnichannel chat and connects to external tools using MCP.
Summary generated from the project's README
About ChatterMate MCP server
Automating customer support through omnichannel chat and intelligent AI agents is what ChatterMate enables for e-commerce, SaaS, and help desk teams. This open-source platform provides a shared inbox where AI handles common queries from a knowledge base and transfers complex issues to human agents. It supports communication across various channels including WhatsApp, Instagram, Messenger, Telegram, Slack, SMS, and email, as well as a web widget for websites and Shopify stores.
Users can connect the platform to their existing infrastructure using the Model Context Protocol to allow agents to query APIs or look up orders mid-conversation. The system is model-agnostic, allowing you to bring your own API keys for providers like OpenAI, Anthropic Claude, Google Gemini, Mistral, xAI Grok, DeepSeek, or Groq. You can deploy it via a free hosted service or self-host the platform to maintain control over your data and workflows.
What you can do with it
- 1Automate omnichannel customer inquiriesAn AI agent can answer customer questions 24/7 across multiple platforms like WhatsApp and Instagram using your specific knowledge base.
- 2Connect AI to internal business toolsA developer can allow the support agent to perform real-time actions like looking up order statuses by connecting MCP servers.
- 3Capture and sync sales leadsThe platform can automatically collect contact information during a chat and send it directly to a CRM like HubSpot.
- 4Automate technical issue investigationThe system can automatically triage support tickets and perform root-cause analysis by gathering evidence from databases.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablechattermate/chattermate.chat is public and reachablegithub.com/chattermate/chattermate.chat ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-12github.com/chattermate/chattermate.chat ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHubOpen-source AI customer support platform with human handoff. ChatterMate is an open source AI chatbot that answers customers 24/7 from your knowledge base, hands complex conversations to your team in one shared inbox, and works across your website, WhatsApp, Instagram, Messenger, Telegram, Slack, SMS and email (omnichannel). It captures and qualifies leads, generates an AI help center, calls your tools over MCP, and runs on the model you choose (OpenAI, Anthropic Claude, Google Gemini, Mistral, xAI Grok, DeepSeek, Groq or any OpenAI-compatible endpoint). Use the free hosted service or self-host it as an open-source alternative to Intercom, Zendesk, Chatbase and Chatwoot.
Documentation | Live Demo | Free Signup | Shopify App | WordPress Plugin
Works with the tools you already use
Channels —
Commerce & CMS —
CRM & tickets —
Data & tools for the AI —
AI models —
Every integration ships in the open-source codebase — see Integrations for setup.
Table of Contents
- Why ChatterMate?
- Features
- Integrations
- ChatterMate vs. Intercom, Zendesk & Chatbase
- Screenshots
- Quick Start
- Installation
- Deployment
- FAQ
- Roadmap
- Contributing
- Support
Why ChatterMate?
ChatterMate is a no-code AI customer support platform that enables businesses to provide 24/7 customer service through intelligent AI agents. Your AI chatbot handles common queries, escalates complex issues to human agents when needed, and continuously learns from your knowledge base. Integrate the chat widget on any website — or any Shopify store — with a single line of code.
Perfect for: e-commerce and Shopify stores, SaaS companies, help desks, customer success teams, and any business looking to automate customer support while maintaining a human touch.

Features
Core Features
| Feature | Description |
|---|---|
| 🤝 Smart Human Handoff | Intelligent AI-to-human transfer with business hours awareness, real-time availability detection, and context-aware escalation messages. |
| 🛍️ Shopify & E-commerce Support | Native Shopify integration (App Store listing) — answer order, shipping, and product questions from store data. Works for any online store. |
| 📝 WordPress Plugin | Official WordPress plugin (download) — add the chat widget to any WordPress site by entering your Widget ID. No theme edits required. |
| 🧠 Multi-Model AI Support | Choose your AI provider — OpenAI, Anthropic (Claude), Google Gemini, Mistral, xAI (Grok), DeepSeek, and Groq — with your own API key, or enter a custom model ID. Switch providers anytime without code changes. |
| 💬 Ask Anything Mode | Let visitors start conversations instantly — no signup or email required. Perfect for Q&A, documentation assistants, and exploratory chat experiences. |
| 📎 File Attachments | Customers can share images, PDFs, Word docs, spreadsheets, and more directly in chat. Secure uploads with S3 storage and magic byte validation. |
| 🌍 Auto Translation | Multilingual support with configurable default language per workflow. Serve customers globally in their preferred language. |
| 🎫 AI Ticketing | Native tickets raised straight from chat, then auto-triaged and investigated by AI — it forms hypotheses, gathers evidence from your observability tools and databases, and writes a root-cause analysis you can read and audit. See AI Ticketing. |
| 🔗 Jira Ticket Creation | Create and manage Jira tickets directly from chat conversations. OAuth 2.0 secure integration with automatic ticket tracking. Native tickets can also escalate to Jira one-way. |
| 🔑 Widget Authentication | Token-based security for embedded widgets. Support both public Q&A and private authenticated conversations. |
| 💼 Slack Integration | Connect your Slack workspace for internal product support. Teams get AI-powered assistance directly in Slack channels. |
| 🧩 Visual Workflow Builder | Design conversation flows with a drag-and-drop interface. Branching logic, conditional responses, and multi-step workflows without coding. |
| 🎯 AI Lead Capture | The agent collects name, email, phone and company during the conversation, with consent, writes a qualification summary, and pushes the lead to HubSpot or Pipedrive. See Lead Capture. |
| 📚 AI-Generated Help Center | Turn your knowledge base into a public, SEO-friendly help center — FAQs are generated by AI, reviewed, then published, with AI search and the chat assistant built in. See Help Center. |
| 🧰 MCP Tools | Connect any Model Context Protocol (MCP) server so the agent can look up orders, query APIs and call your tools mid-conversation. See MCP. |
| 🎨 Chat Customization | Pick a chat design, set brand colours, avatar, welcome message and quick actions, with a live preview. See Customization. |
Workflow Builder
Build sophisticated conversation flows visually with our intuitive workflow builder:
Workflow Builder Features:
- Drag-and-drop nodes - AI responses, human handoff, conditions, and more
- Branching logic - Create different paths based on user input
- Node types - Start, AI Response, Human Transfer, Condition, End nodes
- Real-time preview - Test workflows before deploying
- Version control - Save and restore workflow versions
AI Lead Capture
Turn support and sales chats into qualified leads without a form. When a visitor shows buying intent, the agent asks for their details in the conversation (with a consent step you can switch on), records them on the person's profile, and writes an AI qualification summary so your team knows who to call first.
- One profile per person across the web widget, WhatsApp, Instagram, Messenger, Telegram and email, moving from Visitor → Lead → Customer
- Configurable fields per agent — name, email, phone, company, or your own
- Source tracking — the page or channel each lead came from
- CRM sync — push leads to HubSpot or Pipedrive
- People view to filter, search and follow up on every lead
AI-Generated Help Center
Point ChatterMate at your website, docs or PDFs and it generates FAQ articles with AI, grouped into topics and linked back to their source. You review and publish them as a public knowledge base / help center — no separate help desk tool needed.
- AI-generated FAQs from your knowledge base, or import an existing help center
- Review before publishing — edit, regenerate, draft or publish each article
- SEO-friendly public pages with per-article titles and meta descriptions
- AI search over your articles, and the chat assistant on every page for anything they don't cover
- Your branding — name, logo, colour and header links

Connect Any MCP Server
Give your AI agent real tools. Add any Model Context Protocol (MCP) server — over HTTP, SSE or stdio — and the agent can call it mid-conversation to look up an order, check stock, query an internal API or search logs. Add per-connector guidance so the agent knows when to use each tool, and test the connection from the dashboard.
ChatterMate is also manageable over MCP itself — see the MCP server docs.
Bring Your Own Model
Run your agents on the model you trust: OpenAI, Anthropic Claude, Google Gemini, Mistral, xAI Grok, DeepSeek, Groq, or any OpenAI-compatible endpoint (self-hosted, OpenRouter, Ollama…). Use your own API key, pick from the list or enter any model ID, and switch providers at any time without code changes.
Chat Widget Customization
Make the chat widget look like part of your site: choose a chat design, set colours, avatar, fonts, welcome title and message, starter questions and quick actions, and see every change in the live preview before you save.
AI Ticketing
Tickets are raised from a conversation (by the AI or an agent), or from an alert webhook. Each one is triaged, deduplicated against open tickets, and then investigated: the AI proposes hypotheses, tests each against your connected tools, and records every query it ran as evidence you can inspect. The result is a versioned root-cause analysis with citations back to that evidence.
Connect read-only investigation sources under Settings → Ticketing: observability platforms via MCP (Grafana, Elasticsearch, Sentry, CloudWatch, or any MCP server), and optionally a guardrailed SQL connector (Postgres/MySQL, direct or over an SSH tunnel).
Autonomy is staged, and you choose the level:
| Level | The AI can |
|---|---|
| L1 | Investigate and document only |
| L2 | Also propose a resolution — a human approves or rejects it |
| L3 | Also message the customer and close the ticket, behind confidence guards |
Guardrails
The investigation agent reads untrusted customer text and holds live tool access, so the limits are enforced in code, not by prompting:
- The SQL connector is read-only, structurally. Queries are parsed to an AST
and rejected unless they are a single plain
SELECT. Writes, DDL, multiple statements, and a denylist of dangerous functions cannot pass. Table access is restricted to an allowlist you pick, enforced through CTEs, joins and subqueries. ALIMITis forced. Comments are stripped before execution, since MySQL executes/*! */. The connection itself is a read-only transaction with a statement timeout — a second, independent barrier. - Columns you mask are never readable. Masked columns are blocked from being
referenced anywhere, including in a
WHEREclause, so their values can't be probed; whole-row tricks that would smuggle them out (to_jsonb(t),t::text,to_json(t.*)) are blocked too, and results are masked again on the way back. - Nothing is written to your database, ever. Approving an AI proposal records the decision — any change to your systems is made by your team.
- Row-level scoping keeps one customer's data out of another's ticket. Mark
the column that identifies the customer on any table holding per-customer rows
(
orders → customer_email). Queries against it are rewritten to read only the ticket customer's own rows — the AI cannot widen that, because the filter is applied to the table it selects from, not to a condition it could write around. A ticket with no known customer cannot query a scoped table at all, rather than falling back to reading everything. Tables you leave unscoped (products, error codes) stay fully readable. - Cross-customer isolation on outbound messages. Identifiers belonging to anyone other than the recipient are stripped from every message sent to a customer, and other customers' tickets are redacted before they reach the model — a second line behind row scoping.
- Every query is audited. Each attempt is logged with the SQL and outcome; returned rows are deliberately never stored.
When you connect a database: set a row-scope column for every table that holds per-customer rows — it's the control that stops the agent reading across customers, and it's off until you set it. Point the connector at a service account restricted to what support genuinely needs, and prefer views that are already scoped.
Platform Features
| Feature | Description |
|---|---|
| 🤖 AI-Powered Responses | Context-aware AI with conversation memory across sessions |
| 📊 Analytics Dashboard | Real-time monitoring, conversation insights, and performance metrics |
| 📚 Knowledge Base Training | Train your AI with domain-specific knowledge and FAQs |
| 🎨 Custom Theming | Fully customizable chat widget to match your brand |
| 🔐 Role-Based Access Control | Granular permissions for team members |
| 🌐 Open Source & Self-Hosted | Full control over your data with self-hosting option |
Single sign-on
Team members can sign in to the dashboard with Google or Microsoft
(OpenID Connect) once an operator sets the provider's client id and secret
(GOOGLE_SSO_CLIENT_ID / MICROSOFT_SSO_CLIENT_ID and secrets, see
backend/.env.example). SSO signs in existing team members and never creates
accounts: the first sign-in links the provider account to the team member with
the same e-mail address, when the provider vouches for that address, and later
sign-ins use that link. Each team member links at most one account per
provider. Changing a team member's address drops their links (changing your own
needs your current password), and an admin can list and unlink them
(GET/DELETE /api/v1/users/{id}/identities).
A team member without a profile picture gets the provider's photo. Register
{BACKEND_URL}/api/v1/sso/<provider>/callback as the redirect URI with the
provider. For Microsoft on the default common tenant, add the optional email
and xms_edov claims to the app's ID token (Entra ID → Token configuration):
a work or school account links only when xms_edov says Microsoft verified its
domain, so without it those accounts can't sign in. Step-by-step setup for both
providers is in the single sign-on docs.
With PASSWORD_LOGIN_ENABLED=false the login page leads with SSO, and
a team member's password stops working once they have signed in with SSO, and
their password sessions end (it applies install-wide, so it is meant for
single-organization installs).
Integrations
All of these are free and included — every adapter ships in the open-source codebase, and self-hosting unlocks the lot.
What some of them need is your own credentials, because the integration talks to your account on someone else's platform:
- You create the app — a developer app registered once per ChatterMate
install, with its keys in the backend
.env. - You paste a token — no app to create; you generate a token or API key on the provider's side and enter it in the ChatterMate UI when connecting.
Both are marked in the tables below.
Messaging channels — one shared inbox for all of them, with the same AI answers, human handoff, and workflows as the web widget:
| Channel | Setup | Notes |
|---|---|---|
| Your Meta app | WhatsApp Business Cloud API — templates, media, and 24-hour session handling | |
| 💬 Facebook Messenger | Your Meta app | Meta Messenger Platform, connected per Page |
| Your Meta app | Instagram Direct Messages via the Meta Graph API | |
| 🔷 Slack | Your Slack app | Your Slack workspace as an internal support channel |
| ✈️ Telegram | Paste a token | Bot token from BotFather — no app registration |
| 🟢 LINE | Paste a token | LINE Messaging API channel credentials |
| 📱 SMS | Paste a token | Your Twilio, Vonage, Plivo, or MessageBird account keys |
| Paste a token | Your mailbox credentials; inbound email becomes a conversation and replies go back over email | |
| 🌐 Web widget | Nothing | The embeddable widget, plus Shopify and WordPress surfaces |
CRM — push captured leads and contacts out to your sales stack:
| Integration | Setup | Notes |
|---|---|---|
| 🟠 HubSpot | Your HubSpot app | Contacts and leads pushed from chat, OAuth connected |
| 🟩 Pipedrive | Your Pipedrive app | Persons and leads pushed from chat, OAuth connected |
Synced by the crm_sync worker — see Background processors.
Tickets, data and tooling
| Integration | Setup | Notes |
|---|---|---|
| 🔗 Jira | Your Atlassian app | Create and track Jira issues from chat (OAuth 2.0); native tickets can escalate to Jira |
| 🛍️ Shopify | Your Shopify app | Order, shipping, and product answers from live store data |
| 🧰 MCP servers | Paste a token | Any MCP server as an agent tool — Grafana, Elasticsearch, Sentry, CloudWatch, and others |
| 🗄️ SQL connectors | Paste a token | Read-only, guardrailed PostgreSQL and MySQL access for AI ticket investigation |
| 🪝 Webhooks | Nothing | Outbound ticket webhooks to drive your own automations |
Adding a channel means implementing one adapter in
backend/app/channels/against the sharedChannelAdapterbase — the inbox, AI, and handoff come for free. CRM adapters follow the same shape inbackend/app/crm/.
ChatterMate MCP questions
What channels does ChatterMate support?
ChatterMate works across your website, WhatsApp, Instagram, Messenger, Telegram, Slack, SMS, and email.
Can I use my own AI models?
Yes, you can bring your own LLM by using your own API keys for providers like OpenAI, Anthropic, Google Gemini, and more.
How does it integrate with other tools?
It connects to various platforms including Shopify, WordPress, HubSpot, Pipedrive, and Jira, and can call tools via the Model Context Protocol (MCP).