Chatway Live Chat
io.github.Chatway-app由 GitHub 帳號 Chatway-app 發布v1.3.3更新於 Nov 7, 2026
Connect ChatGPT, Claude & AI agents to Chatway live chat to reply to customers and manage tickets
概覽
遠端 MCP 伺服器,讓 AI 助理讀取並管理 Chatway 線上客服對話、工單、聯絡人與知識庫文章。
- 可以做什麼
- 列出、搜尋並讀取對話與訊息紀錄,將對話標記為已解決或重新開啟
- 回覆客戶,並為團隊新增內部備註
- 列出客服人員、指派對話,並在 AI 與真人客服之間交接
- 搜尋聯絡人並查看訪客詳細資料
- 在知識庫中搜尋說明文章
- 為對話加上標籤、更新聯絡人自訂資料(付費方案),並取得或驗證網站安裝腳本
- 適用情境
- 當客戶支援使用 Chatway,而你想讓助理分流未解決的對話、草擬或送出回覆、彙整客戶問題、把工單轉給合適的客服,或依說明文章作答時適用。也適合搭配程式編寫助理在網站上安裝並驗證 Chatway 小工具。
- 執行需求
- 遠端 Streamable HTTP 端點,無需安裝或自行架設。需要 Chatway 帳號(免費方案即可,部分工具需付費方案),並以工作區擁有者或管理員身分完成 OAuth 2.1 登入。登入時需選擇要分享哪些小工具、頻道與權限。
安裝
在 SourceWeft 中
- 開啟 儀表板中的 Chatway Live Chat,將其新增到工作區。
- 連接 所需的 憑證。它們會被私密儲存,絕不會在公開頁面上展示。
- 為需要使用其工具的對話啟用該服務。
Web executable,透過 Streamable HTTP。 遠端服務在工作區中設定後即可從網頁執行環境執行。
交給代理程式安裝
把這段提示詞貼上到 Claude Code、Codex、Cursor 等代理程式,或 SourceWeft 對話中。代理程式會讀取該伺服器的安裝指南,先向你展示發布者、端點和登入方式,經你同意後再把它新增到你的客戶端。金鑰不會寫進任何命令。
閱讀 https://sourceweft.com/mcp/io.github.Chatway-app%2Fchatway/install.md,並按其中說明把這個 MCP 伺服器接入我的客戶端。修改任何設定之前,先告訴我它的發布者、端點和登入方式,等我確認。不要把 API 金鑰寫進命令或我會提交的檔案;問我金鑰應從哪裡取得。其他 MCP 客戶端
把它新增到你客戶端的 mcpServers 設定中。 客戶端會在首次連接時完成 OAuth 登入。
{
"mcpServers": {
"chatway-live-chat": {
"type": "http",
"url": "https://mcp.chatway.app/mcp"
}
}
}README
Chatway MCP Server: Connect AI Agents to Customer Support
Let ChatGPT, Claude, Gemini, Cursor, and other AI agents read and manage your Chatway live chat conversations, support tickets, and contacts.
Create a free Chatway account · Setup guides · Documentation · Chatway on the Official MCP Registry
The Chatway MCP server connects AI agents to your Chatway inbox over the Model Context Protocol. Ask your assistant to find conversations, summarize customer issues, reply to customers, assign agents, add private notes, search your knowledge base, and organize contacts, all in plain language. Pair it with Chatway webhooks to trigger your agents and workflows the moment a new message or event happens.
Available on every Chatway plan, including Free. There is nothing to install or host and no API key to manage: add the server URL to your AI client and sign in with your Chatway account.
Chatway Live Chat is an AI-powered customer support platform for live chat, AI chatbots, WhatsApp, Facebook Messenger, Instagram, and more, with a shared team inbox, automation, knowledge base, and mobile apps for iOS and Android.
[ChatGPT connected to Chatway through the Chatway MCP server, managing live chat conversations]
Chatway MCP in action: Connect ChatGPT and other AI agents to your website live chat to manage customer conversations, reply to messages, and access your knowledge base. Use webhooks to trigger real-time workflows when new messages or events occur, allowing your AI agents to respond faster and automate customer support without constantly checking for updates.
Quick start
- Create a free Chatway account if you don't have one.
- Add
https://mcp.chatway.app/mcpas a custom (remote) MCP connector in your AI client. - Sign in to Chatway when prompted and choose the widgets, channels, and permissions to share.
- Ask your assistant something like "Show my latest unresolved Chatway conversations."
Supported clients
Tested by Chatway
Works with any client that supports remote MCP servers with OAuth
Codex, Grok, Devin, OpenCode, and others. See the step-by-step setup guides for each client.
What you can do
See the tool reference for every tool and its parameters.
Agent Skills: Ready-to-use AI Customer Support Workflows
Make your AI agents more powerful with Chatway Agent Skills, ready-to-use workflows that help ChatGPT, Claude, Cursor, Codex, and other compatible AI agents manage customer support through Chatway.
While the Chatway MCP Server gives AI agents access to your customer conversations, contacts, and support tools, Agent Skills teach them how to use those tools to complete specific customer support tasks.
Available Agent Skills
Install Chatway Agent Skills
Install the skills directly from the official Chatway GitHub repository:
Or install a specific skill:
Then connect your AI agent to the Chatway MCP server:
https://mcp.chatway.app/mcp
Agent Skills and MCP connections are separate. Your AI client must support installing or loading Agent Skills, and you must separately authorize the Chatway MCP connection. Skill availability and installation methods vary by client.
Example Prompts
Once installed, ask your AI agent:
- "Review my unresolved Chatway conversations and prioritize the ones needing attention."
- "Draft responses for customers waiting for a reply, but don't send them yet."
- "Find conversations that require human support and recommend who should handle them."
- "Summarize this customer's previous conversations and outstanding issues."
- "Analyze this week's support conversations and identify the most common customer problems."
- "Install Chatway live chat and AI chatbot on my website."
Why use Chatway Agent Skills?
- Save time: Turn complex customer support workflows into simple natural-language requests.
- Get better AI responses: Help agents follow structured workflows using actual conversation history and knowledge-base content.
- Keep humans in control: Review proposed responses and approve sensitive actions.
- Work with your favorite AI agents: Use compatible Agent Skills clients alongside Chatway MCP.
- Start for free: Chatway MCP is available on all plans, including Free. Some capabilities require a paid plan.
Explore Chatway Agent Skills | Create a Free Chatway Account
Real-time events with webhooks
MCP lets your agent act in Chatway when you ask it to. Chatway webhooks work the other way: Chatway sends a signed HTTPS request to your endpoint as soon as something happens, such as a new visitor message, a new conversation, or a new contact. Use webhooks to wake up your AI agent or automation (n8n, Zapier, Make, or your own server) instead of polling, then let it read and reply through MCP or the Chatway API.
Webhooks are available on all plans, including Free. Set them up in Chatway under Settings → Developer Tools → Webhooks.
Example: Automatically answer customer support conversations
Combine Chatway webhooks with MCP to build an AI-powered customer support workflow that responds to incoming messages in real time.
Here's how it works:
- Customer sends a message: A visitor sends a question through the Chatway website live chat widget.
- Webhook triggers the workflow: Chatway instantly sends a webhook notification to your AI agent or automation platform (such as n8n, Make, Zapier, or your own server).
- AI agent retrieves context: The agent uses Chatway MCP to read the conversation history, access customer information, and search your knowledge base for accurate answers.
- AI agent responds automatically: The agent generates a relevant response and, when configured and authorized, automatically sends it to the customer through Chatway, without requiring human intervention. Alternatively, the agent can prepare a draft for human review and approval.
- Support team stays informed: The entire conversation remains accessible in the Chatway shared inbox and mobile apps for iOS and Android. Human agents can take over whenever needed.
Fully automated customer support: With Chatway webhooks and MCP, you can connect external AI agents to automatically handle incoming customer messages, answer questions, and manage support conversations 24/7.
Note: MCP provides access to Chatway's tools and data, while webhooks notify your automation when events occur. Automatic responses require an external AI agent or workflow runner configured to process incoming events and execute MCP actions. MCP alone does not continuously monitor conversations.
Learn more about Chatway webhooks | Explore Chatway MCP
Example prompts
- "Summarize today's unresolved support conversations."
- "Find customers waiting more than 2 hours for a reply."
- "Draft replies for all pricing questions, but don't send them yet."
- "Assign today's billing conversations to Alex."
- "Find conversations about refund issues this week."
- "Search the knowledge base for our refund policy and draft an answer for this customer."
- "Summarize the conversation with [email protected] and add a private note with next steps."
- "Tag this visitor as
vipand resolve the conversation." - "Install Chatway on this website."
Install Chatway on your website with a coding agent
Coding agents such as Claude Code, Codex, Cursor, and Replit can add Chatway to your website without you copying the script from the dashboard. For step-by-step help with your framework or platform (Shopify, Wix, WordPress, Webflow, Squarespace and more), add the Install Chatway skill:
Then:
- Connect the Chatway MCP server and approve access. If you do not have an account, sign up at chatway.app first.
- Ask the agent: "Install Chatway on this website."
- The agent calls
get-installation-codeand gets your account'sscript_html. The installation ID identifies your Chatway account, not a widget, so no widget needs to be chosen. - The agent checks that the script is not already on the site, then adds it once to the global
layout, just before
</body>. The agent edits your files; the Chatway server does not. - After you deploy, the agent calls
get-installation-statuswith your domain.widget_operational(visitors have started conversations there) andscript_detected(the script is in the page HTML) confirm the installation. Single-page apps that add scripts after the page loads returnunverified_spa: open the site, start a test chat, then check again.
The same code is available over the REST API with GET /installation/code. See the
tool reference.
Client setup
ChatGPT
Connect Chatway to ChatGPT to search customer conversations, manage support tickets, access contacts, and work with your knowledge base directly from ChatGPT.
- Open ChatGPT on the web and go to Settings → Apps.
- Enable Developer mode (requires an eligible ChatGPT plan).
- Open Browse Plugins, click +, and create a custom MCP app named
Chatway. - Enter
https://mcp.chatway.app/mcpas the MCP server URL and select OAuth authentication. Leave the Client ID and Client Secret empty. - Sign in to Chatway and authorize the widgets, channels, and permissions you want ChatGPT to access.
- Start a conversation, enable the Chatway app, and ask: "Show my latest unresolved Chatway conversations."
Available tools and write capabilities depend on your ChatGPT plan and workspace settings.
Claude
Connect Chatway to Claude to manage customer conversations, draft or send replies, assign support tickets, and search your knowledge base.
- Open Claude and go to Customize → Connectors.
- Click + Add → Add custom connector.
- Enter
Chatwayas the name and pastehttps://mcp.chatway.app/mcpas the remote MCP server URL. - Follow the connector setup flow, using automatic OAuth registration when prompted. No API key or manually created client credentials are required.
- Click Connect, sign in to Chatway, and authorize the widgets, channels, and permissions Claude can access.
- Ask Claude: "Summarize my latest Chatway support conversations."
For Claude Team or Enterprise, an organization owner may need to add the connector first. The connection also works with Claude Desktop and other supported Claude clients.
Claude Code: Run claude mcp add --transport http --scope user chatway "https://mcp.chatway.app/mcp", then authenticate using /mcp.
Gemini CLI
Install the Chatway extension in Gemini CLI:
Then start gemini and sign in to Chatway:
A browser window opens to sign in to Chatway and approve access. Once connected, /mcp list
shows the chatway server and its tools.
Gemini CLI receives the OAuth callback on localhost, so sign in on the same machine where
Gemini CLI is running.
Cursor
Install the Chatway plugin from the Cursor Marketplace. The plugin adds the Chatway MCP server and a Chatway rule that guides the agent to confirm before sending replies or making destructive changes.
- Install the plugin, then open Cursor Settings → MCP (or Tools & MCP).
- Next to the
chatwayserver, click Connect (or Needs login) and sign in to Chatway. - Once connected, the Chatway tools are listed under the server and available to the agent.
If you previously added Chatway manually in ~/.cursor/mcp.json, remove that entry so only the
plugin's server is used.
GitHub Copilot (VS Code)
Chatway works with GitHub Copilot Chat in VS Code in Agent mode.
- Open the Command Palette (
Cmd/Ctrl + Shift + P) and run MCP: Add Server. - Choose HTTP, then Workspace or Global.
- Paste
https://mcp.chatway.app/mcpand name the serverchatway. - Trust the server when VS Code asks, then sign in to Chatway in the browser window that opens.
- In Copilot Chat, switch to Agent mode. The Chatway tools appear in the tools picker.
Or add it to mcp.json yourself (.vscode/mcp.json for a workspace, or MCP: Open User
Configuration for all workspaces):
The root key is servers (not mcpServers), and "type": "http" is required.
One-click install from the GitHub MCP Registry (searching
@mcp Chatwayin VS Code Extensions) will be available once Chatway is accepted into the curated GitHub MCP Registry. Until then, use the steps above.
Cline
Ask Cline: "Set up the Chatway MCP server from https://github.com/Chatway-app/chatway-mcp-server". Cline follows llms-install.md to add the server.
To add it yourself, open MCP Servers → Remote Servers, enter chatway as the name, paste
https://mcp.chatway.app/mcp, choose Streamable HTTP, click Add Server, and sign in to
Chatway when prompted. Or add this to Cline's MCP settings JSON:
"type": "streamableHttp" is required. Without it, Cline uses the legacy SSE transport and the
connection fails with HTTP 400.
goose
goose Desktop: open Extensions → Add custom extension, enter Chatway as the name, choose
Streamable HTTP, paste https://mcp.chatway.app/mcp as the endpoint, and click Add. A
browser window opens so you can sign in to Chatway and choose what goose can access.
Or paste this install link into your browser to add it in one step:
goose CLI: run goose configure, choose Add Extension → Remote Extension (Streamable HTTP),
enter chatway as the name and https://mcp.chatway.app/mcp as the endpoint. Leave the request
headers empty: goose opens the Chatway sign-in in your browser the first time it connects.
Then try: "List my 5 most recent unresolved Chatway conversations."
Grok Build
This repository is a Grok Build plugin: it bundles the Chatway MCP server (.mcp.json) and the
Agent Skills in skills/.
Open /plugins in Grok Build, go to the Marketplace tab, and install chatway. Then open
/mcps, select chatway, and press i to sign in to Chatway in your browser.
To add only the MCP server:
Grok opens the Chatway sign-in the first time it uses the server and stores the token in
~/.grok/mcp_credentials.json. Then try: "List my 5 most recent unresolved Chatway conversations."
Replit
Click Add to Replit above, then Test & save and sign in to Chatway. To add it manually,
open replit.com/integrations, choose MCP → + Add MCP server,
enter Chatway as the display name and https://mcp.chatway.app/mcp as the server URL.
The connection is saved to your Replit account, so Replit Agent can use it in any app. Try: "Use Chatway to list my 5 most recent unresolved conversations." To react to new messages in an app you build on Replit, pair it with Chatway webhooks.
Other clients
Add https://mcp.chatway.app/mcp as a remote MCP server using the Streamable HTTP transport and
complete the Chatway sign-in. Client-specific steps for Codex, Grok, and others are in the
setup guides.
Troubleshooting
- Tools missing or
401errors: reconnect or re-authenticate thechatwayserver in your client. Tokens are revoked when the connection is disconnected in Chatway. - Can't approve access: only Chatway workspace owners and admins can connect an assistant.
- A tool isn't available: that permission wasn't granted, or the tool requires a paid plan. Change permissions under AI → External AI in Chatway, or reconnect.
Security and data protection
- You choose what to share. During sign-in you pick which widgets, channels, and permissions the connection gets. Tools without a granted permission are not offered to the assistant.
- Role-checked. Only workspace owners and admins can connect an assistant, and every action is limited by the authorizing user's Chatway role.
- Server-side confirmation for bulk actions. Resolving, reopening or assigning more than one conversation is not performed on the first call. The server returns a one-time confirmation token tied to the exact request, and the action only runs when the tool is called again with that token. Other actions run straight away within the permissions you granted, so an agent can work unattended. Your AI client may still ask you to approve a tool call on its own.
- No passwords or API keys shared. Connections use OAuth 2.1; your AI client never sees your Chatway password.
- Revocable at any time. Owners and admins can change permissions or disconnect from AI → External AI in Chatway; disconnecting revokes the token immediately.
- Hosted by Chatway. The server runs on Chatway's infrastructure over HTTPS. This repository contains no server code or secrets.
Details: Permissions & access · Manage connections · Security policy · Privacy policy · Terms of service
About this repository
This repository holds Chatway's public listings for the official MCP Registry, the Gemini CLI extension gallery, the Cursor Marketplace, and the Cline MCP Marketplace: manifests, logos, and setup guides. Maintainers: see MAINTAINING.md.
Support
Visit the Chatway Help Center or contact the support team through live chat on chatway.app.
License
The manifests, configuration, and documentation in this repository are available under the MIT License. The Chatway name and logos are trademarks of Chatway and are not covered by this license. The hosted Chatway service and MCP server are proprietary.
來源:README.md,提交 6686a36
工具
0版本歷史
6- v1.3.3最新Nov 7, 2026
- v1.3.0Nov 6, 2026
- v1.2.2Oct 26, 2026
- v1.1.0Oct 20, 2026
- v1.0.3Oct 13, 2026
- v1.0.2Oct 12, 2026
評分
已滿足適用檢查項的 85%
該評分衡量此收錄條目的完整程度,依據註冊中心條目、端點與公開來源計算得出。它不是對伺服器本身的評價,也不表示伺服器安全或目前可用:可連通性是帶日期的檢查結果,評分也不包含安裝量或使用者評價。
- 近 30 天內可連通必需25 分
- 已宣告認證方式必需10 分
- 已列出工具15 分
請讓伺服器在無憑證時回應 tools/list,或發布可在沙箱中無憑證啟動的軟體包。
- 描述必需15 分
- 圖示10 分
- 文件15 分
- 授權條款10 分