What is the Chaterimo MCP server?

The Chaterimo MCP server exposes 19 tools via the Model Context Protocol, letting Claude AI read and update your chatbot data in real time — including conversations, analytics, products, leads, tickets, and chatbot configuration.

Do I need a developer to set up the MCP integration?

No. The Claude Desktop OAuth flow requires only a Chaterimo login — no coding needed. You paste the endpoint URL into Claude Desktop settings, authorize via OAuth, and you are connected.

What can Claude read from my Chaterimo account via MCP?

Conversations, analytics, products, leads, tickets, sales rules, dashboard stats, and full chatbot configuration including AI model, system prompt, and widget appearance.

Can Claude change my chatbot settings via MCP?

Yes, using the full-access endpoint. Claude can update instructions, AI model, language, widget colors, and messages, and trigger retraining. Use the read-only endpoint (/mcp-readonly/) to prevent writes.

Is the MCP connection secure?

Yes. OAuth connections use short-lived tokens. API key connections use your cha_xxx key — treat it like a password and use read-only keys for shared access.

Chaterimo MCP Server: Connect Claude AI to Your Chatbot (2026)

By David Langr • Updated 2026-05-04 • 10 min read

Use Claude Desktop or Claude Code to read conversations, check analytics, update AI instructions, and manage your chatbot widget — directly from Claude.

Developer API cluster

What is MCP and why it matters for e-commerce

MCP stands for Model Context Protocol — an open standard that lets Claude AI connect to external tools and read live data from your systems in real time. Instead of copying a conversation transcript out of Chaterimo and pasting it into Claude for analysis, you ask Claude directly: "What are the top complaints in my chatbot this week?" Claude calls the Chaterimo MCP tools, fetches the data, and answers immediately.

For e-commerce teams, this changes how you interact with your chatbot data. You no longer need to log into a dashboard to check analytics, read a ticket, or update your AI instructions. You do it inside a Claude conversation using plain English.

Chaterimo's MCP server exposes 19 tools spanning chatbots, conversations, products, tickets, leads, and analytics. Twelve are read-only; seven allow writes. You choose which level of access to grant Claude based on how you connect.

      What you can do with Chaterimo + Claude via MCP
      Ask Claude to summarize recent customer complaints from your chatbot
Have Claude rewrite your AI system prompt in a more concise style
Check analytics trends for the last 30 days without opening a dashboard
Search your product catalog and inspect ticket patterns in one conversation
Switch your chatbot's AI model or language setting with a single instruction

    

Two ways to connect: Claude Desktop (OAuth) vs Claude Code CLI

Chaterimo supports two connection methods. The right one depends on who is connecting and how they work.

OAuth / Claude Desktop — no token required

The OAuth endpoints use the standard MCP authorization flow. You paste a URL into Claude Desktop, authorize with your Chaterimo login, and Claude gets a short-lived access token automatically. No API key to manage, no configuration files to edit.

Full access (read + write): https://www.chaterimo.com/mcp/
Read-only (safe for sharing): https://www.chaterimo.com/mcp-readonly/

Best for: business owners, marketers, support managers — anyone using Claude Desktop who wants to query or manage their chatbot without writing code.

API key / Claude Code CLI — stdio transport

The CLI integration uses FastMCP over stdio. You supply your Chaterimo API key (cha_xxx), available from the Developer API dashboard, and Claude Code connects directly to the MCP server process.

Best for: developers running Claude Code in the terminal who want programmatic access integrated into their workflow.

Legacy tokenized URLs: A third connection method — direct URL with embedded token — is available for advanced integrations and automation pipelines. Contact support if you need this format.

12 things you can read from Claude (read-only tools)

All of the following tools are available on both the full-access and read-only endpoints. They give Claude a complete view of your Chaterimo account without changing anything.

list_chatbots read

List all chatbots in your organisation with their IDs, names, and current AI model.

get_chatbot read

Full chatbot configuration: AI model, system prompt, temperature, widget colors, button labels, and all appearance settings.

list_conversations read

Filter conversations by chatbot, date range, and platform (web, WhatsApp, Facebook Messenger).

get_conversation read

Full conversation transcript with all messages, timestamps, and visitor metadata. Supports pagination for long sessions.

search_products read

Search your synced product catalog by keyword — same hybrid search your chatbot uses to answer product questions.

list_products read

Browse all products synced from your e-commerce platform, including price, description, and availability.

list_tickets read

Filter support tickets by status: new, in_progress, resolved, or closed. Useful for spotting open issue patterns.

get_ticket read

Full ticket details including thread, status history, assigned agent, and linked conversation.

list_leads read

All visitor leads captured by the chat widget — email, name, capture date, and source chatbot.

get_analytics read

Chatbot performance metrics for any period up to 365 days: session count, message volume, platform breakdown, and response rates.

list_sales_rules read

View all active sales agent rules — the triggers and messages your proactive sales widget uses to engage visitors.

get_dashboard read

Dashboard overview: total sessions, active chatbots, platform breakdown, and recent activity summary.

7 actions Claude can take on your chatbot (write tools)

Write tools are only available on the full-access endpoint (/mcp/). They let Claude make real changes to your chatbot configuration — useful when you want to update your setup from inside a Claude conversation without switching tabs.

update_chatbot_instructions write

Rewrite the AI system prompt (max 15,000 characters). Recommended length: 15–40 lines for best results. Changes take effect immediately.

update_chatbot_model write

Switch the AI model (GPT-4o, Claude Sonnet, Gemini Flash, and others) and set temperature between 0.1 and 1.0.

update_chatbot_language write

Set a fixed response language or choose "Automatic" to detect each visitor's language and reply accordingly.

update_chatbot_settings write

Update chatbot name, description, enable or disable the widget, and adjust behavior strictness (how closely it follows knowledge-base sources).

update_widget_appearance write

Set widget colors using hex codes, change position (left, center, right), and adjust bubble style to match your brand.

update_widget_messages write

Edit welcome message, input placeholder text, "thinking" animation label, and submit button label — all visible to your visitors.

retrain_chatbot write

Trigger a full knowledge base retraining job. The chatbot re-indexes all connected sources: products, documents, and custom content.

Step-by-step: Claude Desktop setup (OAuth)

This is the fastest way to connect for non-developers. No configuration files, no API keys — just an OAuth flow.

Open Claude Desktop and navigate to Settings → Integrations (also listed as "MCP Servers" in some versions of the app).
Click "Add MCP Server" and paste one of the following URLs:
- Full access (read + write): https://www.chaterimo.com/mcp/
- Read-only: https://www.chaterimo.com/mcp-readonly/
Claude Desktop opens an OAuth authorization page in your browser. Log in to your Chaterimo account and click Authorize. Claude receives a short-lived token automatically — you do not see or store it.
Claude Desktop now shows "Chaterimo" in your tools list with a green connected indicator. The 19 tools (or 12 read-only tools) are now available.
Start a new conversation and try: "List my chatbots" or "What is my analytics for the last 7 days?" Claude will call the MCP tools and return a formatted answer.

When to use the read-only endpoint

Use /mcp-readonly/ if you want to give Claude access to your data — conversations, analytics, tickets — without allowing it to change any chatbot settings. This is the right choice for team members or analysts who should be able to query data but not modify your live chatbot configuration.

Step-by-step: Claude Code CLI setup

For developers using Claude Code in the terminal, the MCP server runs as a local process over stdio transport. You need your Chaterimo API key from the Developer API dashboard.

Get your API key: Log in to Chaterimo → Developer API → API Keys → Create new key. The key starts with cha_.

Add the following to your .claude/settings.json (or claude_desktop_config.json for Claude Desktop CLI mode):

// .claude/settings.json or claude_desktop_config.json
{
  "mcpServers": {
    "chaterimo": {
      "command": "python",
      "args": ["-m", "developer_api.mcp_server"],
      "env": {
        "CHATERIMO_API_KEY": "cha_your_key_here"
      }
    }
  }
}

Or run the server directly in a terminal session for quick testing:

# Set your key and start the MCP server process
CHATERIMO_API_KEY=cha_xxx python -m developer_api.mcp_server

Once connected, Claude Code can call any of the 19 Chaterimo tools during a session. You will see tool call output inline as Claude fetches data from your account.

Keep your API key private

Your cha_xxx key grants the same access level as your Chaterimo login for API operations. Do not commit it to version control. For shared environments or team contexts where you do not want write access, create a separate read-only API key from the Developer API dashboard and use that instead.

Real use case examples

Here are five concrete Claude prompts that show what the MCP integration looks like in practice. Each prompt triggers specific Chaterimo tools automatically.

"List all my chatbots and tell me which one has the most conversations this month."
Claude calls list_chatbots to get your chatbot IDs, then calls get_analytics for each one with a 30-day range. It compares session counts and returns a ranked summary with the top performer highlighted.
"Show me the last 10 conversations where a customer mentioned 'return'."
Claude calls list_conversations to page through recent sessions, then calls get_conversation on candidates. It filters transcripts for the keyword and presents excerpts with timestamps and chatbot names.
"Rewrite the instructions for chatbot #3 to be more concise and friendly."
Claude calls get_chatbot to read the current system prompt, proposes a revised version in the conversation for your review, then calls update_chatbot_instructions to apply it after you confirm.
"Switch my main chatbot to GPT-4o with temperature 0.3."
Claude calls list_chatbots to identify which chatbot is marked as primary, then calls update_chatbot_model with the target model and temperature value. The change takes effect immediately.
"What are my top support tickets right now and what patterns do you see?"
Claude calls list_tickets filtered to status new and in_progress, then calls get_ticket for each open item. It groups tickets by topic, identifies recurring issues, and summarizes the most common customer problems in a structured report.

Read-only vs full access: which should you use?

The right choice depends on your role and how much you want Claude to be able to change. When in doubt, start with read-only and upgrade when you need it.

	Read-only	Full access
View conversations, analytics, leads	Yes	Yes
Read chatbot config and products	Yes	Yes
Update chatbot instructions	No	Yes
Change AI model or language	No	Yes
Modify widget appearance	No	Yes
Trigger retraining	No	Yes
Endpoint (OAuth)	`/mcp-readonly/`	`/mcp/`
Best for	Analysts, team members	Owners, developers

Rate limits by plan

MCP tool calls count against your Developer API quota. Each tool invocation — whether Claude calls it automatically during a conversation or you trigger it manually — counts as one API call.

Plan	Calls per minute	Calls per day
Free	5	5
Trial	10	100
Paid	30	1,000

All tool calls are logged in your Chaterimo dashboard under Developer API → MCP Sessions. You can review which tools were called, when, and how many calls remain in your daily quota.

Everything in one place

The Chaterimo MCP server connects Claude to your live chatbot data — conversations, analytics, products, tickets, and configuration — without copy-pasting or tab-switching. Whether you use Claude Desktop with OAuth or Claude Code with an API key, the same 19 tools are available to query and manage your chatbot in plain English.

No copy-pasting conversation logs into Claude
Ask natural-language questions about your chatbot performance
Update chatbot instructions from inside a Claude conversation
Works with Claude Desktop (OAuth) and Claude Code (API key)
Read-only endpoint for safe team sharing

Get Started Free Developer API Docs