Connect your AI agent to Quriobot (MCP)
Your AI agent can work in your Quriobot account for you: answer and close live chats, look up responses and contacts, check how a bot is set up, change its settings. Anything you do in control.quriobot.com, the agent can do through the same API, acting as you.
It works with any AI agent that supports the Model Context Protocol (MCP): Claude Code, Claude Desktop, Cursor, VS Code with GitHub Copilot, Gemini CLI, Codex and more. The connector runs on your own computer and needs Node.js 18 or later.
The recommended way to run it is npx -y @quriobot/mcp@latest, which every setup below uses: npx fetches the newest release each time your agent starts the connector, so you get updates without doing anything.
Before you start: your API key
Create an API key in control.quriobot.com: open your Profile, the API keys tab, and add a key. The Quriobot API article shows where to find it.
The key acts as you, in every organisation you are a member of. Give it a name you will recognise, such as “Claude on my laptop”. You can delete it there at any time, and the list shows when it was last used.
Claude Code
In Claude Code, type:
/plugin marketplace add quriobot/quriobot-mcp
/plugin install quriobot@quriobot
Claude Code asks for your API key while installing and keeps it in its credential store, not in a settings file. The plugin runs npx -y @quriobot/mcp@latest as well, so it stays up to date. To change the key later: /plugin configure quriobot@quriobot.
Claude Desktop, Cursor, Gemini CLI and other agents
Add this entry under mcpServers in the agent’s configuration file, with your own key:
- Claude Desktop: Settings → Developer → Edit Config (
claude_desktop_config.json) - Cursor:
~/.cursor/mcp.json - Gemini CLI:
~/.gemini/settings.json
{
"mcpServers": {
"quriobot": {
"command": "npx",
"args": ["-y", "@quriobot/mcp@latest"],
"env": { "QURIOBOT_API_KEY": "your API key" }
}
}
}
Restart the agent afterwards.
VS Code (GitHub Copilot)
Add .vscode/mcp.json to your project. VS Code asks for the key once and stores it securely:
{
"inputs": [{ "type": "promptString", "id": "quriobot-key", "description": "Quriobot API key", "password": true }],
"servers": {
"quriobot": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@quriobot/mcp@latest"],
"env": { "QURIOBOT_API_KEY": "${input:quriobot-key}" }
}
}
}
Codex
Add to ~/.codex/config.toml:
[mcp_servers.quriobot]
command = "npx"
args = ["-y", "@quriobot/mcp@latest"]
env = { QURIOBOT_API_KEY = "your API key" }
Configuration files hold the key in plain text: keep them private and out of version control.
Try it
Ask your agent, for example:
- “Which live chats in my Quriobot organisation are waiting for an answer?”
- “Summarise this week’s responses of my lead bot.”
- “Why does my bot not show on my website? Check its settings.”
- “Draft a reply to the last chat about pricing, but don’t send it.”
Stay in control
The connector has separate tools for reading and for changing things. Most agents ask before every tool call; let yours read freely and keep approving changes yourself. In Claude Code, for example, allow read, find_endpoints and describe_endpoint, and keep write on “ask”.
- A change can reach your customers: sending a chat message, closing a chat, changing a bot.
- Chats and contacts contain what other people wrote. Your agent is told to treat that as information, not as instructions, but approving changes yourself is the safe default.
- Delete the API key in your profile to cut the agent off at once.
The connector is open source: github.com/quriobot/quriobot-mcp.
Quriobot Academy