Phone-a-Friend MCP Server

Phone-a-Friend MCP Server is an MCP server that creates an AI-to-AI consultation system for complex reasoning and problem-solving tasks. It connects a host AI client to external large language models via providers such as OpenAI, OpenRouter, Anthropic, and Google. Software developers and technical teams use it when an assistant requires external expertise, deep multi-step analysis, or long-context reasoning beyond immediate session capabilities. The server enables an AI to package project context, source code files, and custom task instructions to query another model directly. By handling the automated reading of codebase files while filtering out gitignored or binary content, it extracts actionable insights and returns structured results back to the primary agent. Alternatively, it can export compiled context into a markdown master prompt file for manual copy-paste consultations.

Category: AI & LLM Tooling

Tags: llm, multi-agent, openrouter, reasoning

Visit Phone-a-Friend MCP Server

How to install and configure Phone-a-Friend MCP Server

To install and configure Phone-a-Friend MCP Server in Claude Desktop or a compatible MCP client: 1. Ensure you have uv installed on your machine to execute the server package automatically. 2. Open your client configuration file (e.g., claude_desktop_config.json). 3. Add the server under the mcpServers object: json { "mcpServers": { "phone-a-friend": { "command": "uvx", "args": [ "phone-a-friend-mcp-server", "--provider", "openai", "--api-key", "<YOUR_API_KEY>" ] } } } 4. Replace <YOUR_API_KEY> with your actual provider key, or set environment variables like OPENAI_API_KEY or OPENROUTER_API_KEY. 5. Restart your client to access the phone_a_friend and fax_a_friend tools.

What you can do with Phone-a-Friend MCP Server

• Consulting advanced external reasoning models to debug complex algorithms, state management issues, or data structure implementations across multiple source files. * Analyzing extensive codebase context and tracebacks to solve intricate frontend architecture challenges in modern frameworks like React or Vue. * Packaging full repository context and task descriptions into a markdown file using the fax tool for manual multi-agent reviews. * Gaining a second perspective on critical technical decisions by delegating high-stakes system design questions to specialized external models.

Key facts

• https://github.com/abhishekbhakat/phone-a-friend-mcp-server
• AI & LLM Tooling, Communication & Messaging
• llm, multi-agent, openrouter, reasoning

Part of MCP Servers

How do I install Phone-a-Friend MCP Server?

You can run Phone-a-Friend MCP Server using the uvx runner without manual repository cloning. Add the server entry to your MCP client configuration file specifying uvx, the package name phone-a-friend-mcp-server, your selected provider flag, and your API key. After saving the configuration and restarting your client, the consultation tools become available.

What tools does Phone-a-Friend MCP Server provide?

The server provides two primary tools. The phone_a_friend tool makes direct API calls to consult external AI models for critical reasoning and returns actionable insights to the primary AI. The fax_a_friend tool compiles your contextual files and instructions into a single master markdown prompt file for manual copy-paste workflows.

Which model providers are supported?

The server supports multiple backend providers including OpenAI, OpenRouter, Anthropic, and Google. Users can choose default models such as OpenAI o3, Anthropic Claude 4 Opus, or Gemini 2.5 Pro Preview, or specify custom models and custom base URLs through CLI flags and environment variables.

Is Phone-a-Friend MCP Server open source?

Yes, Phone-a-Friend MCP Server is open source and released under the MIT License. The code and documentation are available in its GitHub repository.