# ChatbotX — MCP server card for AI agents

Omnichannel customer messaging, CRM contacts, flows, sequences, broadcasts, and AI automation.

Registry name: `io.chatbotx/mcp` · category: Commerce and payments · version 1.8.1 · updated 2026-11-06

## Connect
- Remote (Claude Code): `claude mcp add --transport http chatbotx https://app.chatbotx.io/mcp/messages`
- Remote (Claude Code): `claude mcp add --transport http chatbotx https://app.chatbotx.io/mcp/sse`

## Facts
- Repository: https://github.com/ChatbotXIO/ChatbotX (1,136 stars)
- Website: https://chatbotx.io

Before you connect a server: read its repository, and give it only the permissions and keys it needs.

## Links
- AgentGid page: https://agentgid.com/mcp/servers/io-chatbotx-mcp/


## ChatbotX: run chat marketing from your AI assistant

ChatbotX is a chat marketing platform for WhatsApp, Messenger, Instagram, Telegram, email and more. This server lets your AI assistant find contacts, send messages, start automated chats and read campaign numbers in your ChatbotX workspace. Every change waits for your yes before it goes through.

### Where to use it
- Manage leads by just asking: Ask your assistant to find new leads, update their details, add tags or put them into a follow-up series, without opening the dashboard.
- Route new chats to the right flow: Your assistant reads a conversation, works out what the customer wants and starts the matching automated chat, or hands it to a person on your team.
- Build and check chat flows: Draft a new automated chat in your code editor, check it against the official format, and publish it only when it passes.
- Weekly campaign numbers: Get a short report on new contacts, blocked contacts, and how your broadcasts, flows and follow-up series performed.
- Connect a self-hosted ChatbotX: If you run ChatbotX on your own server, your assistant starts the helper on your computer and talks to your own instance with a workspace token.
- Clean up contacts before a campaign: Before a big send, have your assistant list contacts without a key tag or field, and fix them one by one with your approval. (idea, not from the official pages)

### How to start
- Add the hosted server to Claude Code. The first use opens a browser where a workspace owner signs in and picks the workspace.

```
claude mcp add chatbotx -t http -s user \
  "https://app.chatbotx.io/mcp/messages"
```

- Ask the assistant to check the workspace first, so it learns the channels, tags and flows it can use.

### License
The code is free to use, change and sell under the MIT license, except one enterprise folder under a commercial license. Using the hosted ChatbotX service is a separate paid or free plan.

### Good to know
- You need a ChatbotX workspace, and only its owner or a super admin can approve the connection.
- The confirmation step is a safety habit, not a lock. The assistant can skip it by sending the confirm flag, so keep an eye on what it sends.
- Sending messages, flows and broadcasts reaches real customers on real channels. Deleting things is possible through the hidden full list of actions.
- A read-only token only sees reading tools, which is a safe way to start.
- Paid cloud plans start at $29 a month; the self-hosted Community plan is $0.

Checked 2026-11-10 against: https://github.com/ChatbotXIO/ChatbotX/blob/d50c84a5cad67917cc294e5aebebb59f028a452c/apps/mcp-server/README.md, https://github.com/ChatbotXIO/ChatbotX/blob/d50c84a5cad67917cc294e5aebebb59f028a452c/apps/mcp-server/src/server/meta-tools.ts, https://github.com/ChatbotXIO/ChatbotX/blob/d50c84a5cad67917cc294e5aebebb59f028a452c/apps/mcp-server/src/server/hitl-gate.ts, https://github.com/ChatbotXIO/ChatbotX/blob/d50c84a5cad67917cc294e5aebebb59f028a452c/apps/mcp-server/src/env.ts, https://github.com/ChatbotXIO/ChatbotX/blob/d50c84a5cad67917cc294e5aebebb59f028a452c/README.md, https://github.com/ChatbotXIO/ChatbotX/blob/d50c84a5cad67917cc294e5aebebb59f028a452c/LICENSE, https://chatbotx.io/docs/mcp/introduction.md, https://chatbotx.io/docs/mcp/tools.md, https://chatbotx.io/pricing, https://registry.npmjs.org/chatbotx-mcp

### Mistakes to avoid
- Don't: Put the workspace token in the connector URL (?workspace_token=...) for Claude Code or Cursor → Do: Use https://app.chatbotx.io/mcp/messages with browser sign-in; keep tokens for the local stdio setup only (Tokens in URLs end up in logs and shell history; the hosted endpoint signs in through the browser.)
- Don't: Repeat a CONFIRMATION_REQUIRED call with _confirm true on your own → Do: Show the proposed action and current state to the user and repeat only after a clear yes (The server does not stop a repeated call; the approval is the agent's job.)
- Don't: search_tools with a whole user request in another language → Do: search_tools with one English action plus one resource, e.g. "add tag to contact" (The search ranks English tool names and descriptions and does not translate.)
- Don't: call_tool with every field wrapped in body, params or input → Do: Pass the fields flat at the top level of arguments, as the tool's inputSchema declares them (The pre-flight check rejects wrapped arguments.)
- Don't: Drop the response of connections_create with provider api → Do: Save secret.token from that response right away (The token is shown once and cannot be fetched again.)

### Questions
- Do I have to paste a secret key into my assistant? No for the hosted server. Claude Code, Cursor and ChatGPT sign in through the browser; a workspace token is only needed for the local or self-hosted setup.
- Why does my assistant not see a delete tool? Deletes and less common actions are left out of the default list. The assistant finds them with search_tools and runs them with call_tool.
- Can I connect one assistant to two workspaces? One connection acts on one workspace, chosen at sign-in. To switch, disconnect and connect again.
- Which Node.js version does the local setup need? Node.js 24 or newer.

---
AgentGid's catalog is curated: every entry was checked against the maker's official pages (registry, repository, license and pricing pages) on the date shown; a license value shows the quote it stands on wherever AgentGid has one. Usage numbers come from AgentGid's own scan of public repositories set up for AI coding agents. Claims by makers are reported, not tested by AgentGid.
