MCP-first · Agentic Web
The MCP server for agentic customer service
Every LoyJoy AI agent uses external MCP servers as tools. And it is itself an MCP server. Make your services available in the agentic web and work with your LoyJoy data directly in the Claude Cowork desktop app (Mac or Windows), in OpenAI ChatGPT (Codex) and in Gemini. No additional cost, no add-on module, from the Professional plan onwards.
What is MCP?
MCP, the Model Context Protocol, is an open standard that connects AI assistants like the Claude Cowork desktop app, OpenAI ChatGPT (Codex) or Gemini with external systems, comparable to a USB-C port for AI. Instead of building every integration from scratch, AI assistants and enterprise systems speak a single, unified protocol.
At LoyJoy every AI agent is also an MCP server. It works in both directions: external AI assistants reach your services through your LoyJoy agent, and your team accesses LoyJoy data directly from Claude Cowork or ChatGPT Codex. Both ways are GDPR-compliant, EU-hosted and governed by granular access control.
Client, outbound, inbound
One protocol, three use cases
MCP connects the LoyJoy platform with the world of AI agents. One protocol, three use cases that reinforce each other.
- ●MCP servers as tools.
- Connect any MCP-enabled system such as ERP, shop, CRM or internal APIs to your LoyJoy agent. The agent uses its tools autonomously, with no custom integration and no code.
- ●Outbound: your services in the agentic web.
- Your customers' personal AI agents access your services, products and processes via your LoyJoy AI agent, securely, in a controlled manner and end-to-end.
- ●Inbound: LoyJoy data directly in Claude Cowork, ChatGPT Codex and more.
- Operator teams connect their tenant to the Claude Cowork desktop app (Mac or Windows), to OpenAI ChatGPT (Codex connector) or to Gemini and work with platform data in natural language, querying analytics, processes and knowledge sources without switching tools.
Live: analytics in Claude Cowork via MCP
Shown in 2 minutes: connect your tenant, ask in natural language, get an interactive infographic and presentation directly from the Claude Cowork desktop app. The setup works the same way with OpenAI ChatGPT (Codex).
Analytics in Claude Cowork via MCP, in 2 minutes
Click to play. The video is only loaded from YouTube after you click (privacy-friendly).
What your team asks in Claude Cowork and ChatGPT Codex
Once your tenant is connected, you work with your platform data in natural language.
Analytics
”How many requests has the customer service agent automated in the past 30 days? Give me a breakdown by process type.”
Management update
”Draft an executive summary for the board: automation rate, top 3 request categories, trend vs. previous quarter.”
Knowledge search
”Search the knowledge base for all articles on cancellation and summarize the key points.”
Process analysis
”Which process steps have the highest drop-off rates? What might be the reasons and what optimizations do you recommend?”
Content creation
”Based on our latest release notes, write a LinkedIn post highlighting the most important new features for enterprise customers.”
Comparison
”Compare the automation rates of our three AI agents and present the results as an infographic table.”
What’s next
BetaComing soon: build AI agents in dialog with Claude Cowork
Through MCP you will soon be able to set up and evolve a LoyJoy AI agent directly in a conversation with the Claude Cowork desktop app. You describe in natural language what the agent should do, Claude Cowork configures processes and knowledge sources in LoyJoy. Currently in beta, fully available at launch.
Tool overview
What your LoyJoy MCP server offers
Once you connect your tenant to the Claude Cowork desktop app, to OpenAI ChatGPT (Codex) or to another MCP client, these tools are available to you.
- Retrieve analytics.
- KPIs per process and sub-process, for any time period.
- Processes & views.
- Browse AI agents, BPMN definitions, home pages and navigation structures.
- Knowledge base.
- Semantic search across your knowledge base.
- Website content.
- Read and create blog posts, cases and releases.
- Emails & files.
- Gmail, Google Drive and connected data sources.
- Web search.
- Brave, DuckDuckGo and Tavily for in-context research.
Key benefits of the LoyJoy MCP server
Future-proof, scalable, GDPR-compliant. MCP brings your business into the agentic web.
- Future-proofing and competitive advantage.
- Position yourself as a pioneer and secure a critical edge in tomorrow's customer communication.
- Seamless integration.
- Enable external AI agents to exchange information smoothly and execute processes directly via your LoyJoy AI agent.
- Efficiency and scalability.
- Automate handling of requests from various AI agents and significantly reduce the workload on your service teams.
- Expanded digital reach.
- Be present and responsive exactly where your future customers operate, via their personal AI assistants.
- Controlled data access.
- Precisely define which information, services and processes external AI agents may access.
- Process automation.
- Execute complex, customer-oriented processes end-to-end in the agentic web, not just answer questions.
Setup in 3 minutes
How to connect your LoyJoy tenant to Claude Cowork or ChatGPT Codex
Shown with the Claude Cowork desktop app (Mac or Windows). The setup works the same way with OpenAI ChatGPT (Codex connector), Gemini, Cursor or any custom MCP client.
- 1
Open settings
In the Claude Cowork desktop app (Mac or Windows), open the settings and navigate to “Connectors”. In OpenAI ChatGPT, do the same under “Connectors”.
- 2
Create a connector
Add a new custom connector, give it a name (e.g. “LoyJoy MCP”), enter
https://app-cloud.loyjoy.com/mcpas the URL and click “Add”. - 3
Log in and get started
Claude Cowork (or the ChatGPT Codex agent) opens the LoyJoy login. Sign in with your passkey as usual and LoyJoy MCP is immediately available. Start with: “Show me all LoyJoy tools.”
Platform comparison
LoyJoy MCP in comparison
MCP is the new standard for AI connectivity. Not every platform is ready for it.
| Capability | LoyJoy | Typical chatbot builder | Contact-center platform |
|---|---|---|---|
| External MCP servers as agent tools (client) | ✓ included | no | no |
| MCP server (outbound) | ✓ included | no | no |
| Analytics in Claude Cowork / ChatGPT Codex (inbound) | ✓ included | no | no |
| Granular access control | ✓ | no | partial |
| EU hosting and GDPR-compliant | ✓ | varies | varies |
| No-code setup | ✓ | limited | no |
| Chat and voice (phone) | ✓ | no | ✓ |
Frequently asked questions about MCP at LoyJoy
-
An MCP server exposes data and functions through the Model Context Protocol so that AI assistants like the Claude Cowork desktop app, OpenAI ChatGPT (Codex) or Gemini can access them in a standardized way. At LoyJoy, every AI agent is also an MCP server.
-
Yes. Every LoyJoy AI agent can connect to any MCP server and use its tools autonomously, for example your ERP, shop system or internal APIs. Access is governed by the same granular permission model.
-
Yes. From the Professional plan onwards you connect your tenant to the Claude Cowork desktop app (Mac or Windows), to OpenAI ChatGPT (Codex connector), to Gemini, to Cursor or to a custom MCP client and query analytics, processes and knowledge sources in natural language.
-
Yes. Access is granularly governed by the LoyJoy roles and permissions model and all tokens can be revoked at any time. The platform is EU-hosted and GDPR-compliant.
-
MCP is included from the Professional plan onwards at no additional cost and without an add-on module. Every AI agent is also an MCP server.
-
In beta, yes. Through MCP you will be able to set up and evolve a LoyJoy AI agent directly in a conversation with the Claude Cowork desktop app. This feature is currently in beta and will be fully available at launch.
Experience MCP in practice
In a personal demo we'll show you how your team will use analytics, knowledge sources and processes directly in the Claude Cowork desktop app or in OpenAI ChatGPT with Codex tomorrow.