# What AI Models Are Integrated with OfficeCLI? Complete MCP Integration Guide > Discover which AI models OfficeCLI integrates with. Learn how the MCP server connects Claude Code, Cursor AI, VS Code Copilot, and LM Studio for seamless document manipulation. - Repository: [OfficeAI/OfficeCLI](https://github.com/iofficeai/OfficeCLI) - Tags: deep-dive - Published: 2026-07-26 --- **OfficeCLI integrates natively with Claude Code, Cursor AI, VS Code Copilot, and LM Studio through a built-in MCP (Model Context Protocol) server that exposes document manipulation commands to these AI agents.** OfficeCLI, the open-source command-line automation tool for Microsoft Office documents, ships with first-class AI agent support via the **Model Context Protocol (MCP)**. According to the iOfficeAI/OfficeCLI repository, this architecture allows leading AI coding assistants to directly create, read, and modify Word, Excel, and PowerPoint files through standardized JSON-RPC channels. ## Supported AI Models and Agents OfficeCLI registers its document manipulation capabilities with four major AI platforms. Each integration uses a dedicated CLI command to start the MCP server and bind it to the specific agent. ### Claude Code (Anthropic) **Claude Code** by Anthropic connects to OfficeCLI through the MCP registration command. When activated, Claude can issue natural language instructions that translate directly to PowerPoint, Word, and Excel operations. Registration command: ```bash officecli mcp claude ``` This command is documented in [`README.md`](https://github.com/iOfficeAI/OfficeCLI/blob/main/README.md) at line 72 and implemented in [`src/CommandMcp.cs`](https://github.com/iOfficeAI/OfficeCLI/blob/main/src/CommandMcp.cs), which handles the parsing and server initialization for the Claude agent. ### Cursor AI **Cursor AI** integrates seamlessly with OfficeCLI's document automation suite. The integration allows Cursor's codebase-aware AI to generate and modify Office files as part of software development workflows. Registration command: ```bash officecli mcp cursor ``` As noted in the repository documentation (line 73), this exposes all OfficeCLI commands to Cursor's agent context. ### VS Code Copilot **VS Code Copilot** users can leverage OfficeCLI directly within their IDE. The integration bridges GitHub's AI coding assistant with document generation tasks, enabling automated report creation from within the editor. Registration command: ```bash officecli mcp vscode ``` The VS Code integration (documented at line 74) utilizes the same MCP server architecture, making OfficeCLI commands available to Copilot's agent mode. ### LM Studio (Local LLMs) **LM Studio** support enables any locally-hosted large language model to control Office documents. This model-agnostic integration works with Llama, Mistral, or any other LLM running through the LM Studio desktop application. Registration command: ```bash officecli mcp lmstudio ``` Documented at line 75 in [`README.md`](https://github.com/iOfficeAI/OfficeCLI/blob/main/README.md), this option is critical for organizations requiring air-gapped or privacy-preserving AI document automation. ## How the MCP Integration Works The **Model Context Protocol** serves as the abstraction layer between OfficeCLI's command interface and AI agent clients. When you run `officecli mcp `, the application starts a JSON-RPC endpoint that translates between the agent's natural language requests and OfficeCLI's native commands. ### Core Implementation Files | File | Function | |------|----------| | [`src/CommandMcp.cs`](https://github.com/iOfficeAI/OfficeCLI/blob/main/src/CommandMcp.cs) | Parses `officecli mcp ` commands and initializes the MCP server | | [`sdk/python/officecli.py`](https://github.com/iOfficeAI/OfficeCLI/blob/main/sdk/python/officecli.py) | Python SDK implementation that agents use to issue commands programmatically | | [`README.md`](https://github.com/iOfficeAI/OfficeCLI/blob/main/README.md) (lines 72-75) | Documents the supported agents and exact registration syntax | The server exposes **create**, **read**, and **modify** operations, allowing AI models to manipulate document structures without understanding the underlying Office file formats. ## Registering AI Agents Step-by-Step To connect an AI agent to OfficeCLI, execute the corresponding registration command in your terminal: 1. **Register Claude Code:** ```bash officecli mcp claude ``` 2. **Register Cursor AI:** ```bash officecli mcp cursor ``` 3. **Register VS Code Copilot:** ```bash officecli mcp vscode ``` 4. **Register LM Studio:** ```bash officecli mcp lmstudio ``` Once registered, the MCP server runs continuously, listening for JSON-RPC requests from the respective AI tool. ## Using AI to Manipulate Office Documents After registration, AI agents invoke OfficeCLI commands through the MCP channel. For example, Claude Code can add a slide to a presentation using natural language that translates to the following command execution: ```bash # AI agent (via MCP) creates a new slide officecli add deck.pptx / --type slide --prop title="Q4 Report" ``` The Python SDK in [`sdk/python/officecli.py`](https://github.com/iOfficeAI/OfficeCLI/blob/main/sdk/python/officecli.py) handles the programmatic interface, while [`src/CommandMcp.cs`](https://github.com/iOfficeAI/OfficeCLI/blob/main/src/CommandMcp.cs) manages the underlying protocol translation. This architecture ensures that any future AI agent supporting the MCP protocol can integrate with OfficeCLI using the same `officecli mcp ` pattern. ## Summary - OfficeCLI ships a built-in **MCP server** that supports **Claude Code**, **Cursor AI**, **VS Code Copilot**, and **LM Studio**. - Registration uses simple commands: `officecli mcp claude`, `officecli mcp cursor`, `officecli mcp vscode`, and `officecli mcp lmstudio`. - The integration layer is implemented in [`src/CommandMcp.cs`](https://github.com/iOfficeAI/OfficeCLI/blob/main/src/CommandMcp.cs) and documented in [`README.md`](https://github.com/iOfficeAI/OfficeCLI/blob/main/README.md) lines 72-75. - AI agents communicate via **JSON-RPC** to execute create, read, and modify operations on Office documents. - The **model-agnostic MCP architecture** ensures compatibility with future AI coding assistants. ## Frequently Asked Questions ### What is the Model Context Protocol in OfficeCLI? The **Model Context Protocol (MCP)** is a standardized interface that allows OfficeCLI to expose its document manipulation commands to AI coding assistants. According to the source code in [`src/CommandMcp.cs`](https://github.com/iOfficeAI/OfficeCLI/blob/main/src/CommandMcp.cs), this protocol translates between JSON-RPC messages from AI agents and native OfficeCLI operations, enabling models like Claude and GPT-4 to edit PowerPoint and Excel files directly. ### Can I use OfficeCLI with local AI models? Yes. OfficeCLI supports **LM Studio** through the `officecli mcp lmstudio` command, which connects the MCP server to any locally-hosted LLM. This integration works with open-source models like Llama, Mistral, or fine-tuned variants, making it ideal for air-gapped environments or privacy-sensitive document processing. ### How do I switch between different AI agents in OfficeCLI? Switching agents requires running the specific registration command for your desired AI tool. Execute `officecli mcp claude` for Anthropic's agent, `officecli mcp cursor` for Cursor AI, `officecli mcp vscode` for GitHub Copilot, or `officecli mcp lmstudio` for local models. Each command reconfigures the MCP server to listen on the appropriate channel for that specific agent. ### Which Office document operations can AI agents perform through OfficeCLI? AI agents can perform **create**, **read**, and **modify** operations on Word, Excel, and PowerPoint files. Through the MCP integration, agents invoke commands like `add`, `set`, and `view` to manipulate slides, worksheets, and document properties. The Python SDK ([`sdk/python/officecli.py`](https://github.com/iOfficeAI/OfficeCLI/blob/main/sdk/python/officecli.py)) provides the programmatic interface that agents use to execute these tasks.