Pipepie is an MCP server tool built with Go, designed to create self-hosted encrypted tunnels. It connects webhooks, AI pipelines, and local development environments, enabling AI agents to interact with external data and tools. This program supports the open standard Model Context Protocol, allowing AI agents direct access to tunnels, requests, and pipeline traces. It provides secure exposure for local services and offers capabilities for AI pipeline tracing and webhook inspection. Pipepie serves developers and AI agents by bridging local environments with internet services and AI workflows.

What does Pipepie connect to?

Pipepie functions as a self-hosted encrypted tunnel, linking webhooks, AI pipelines, and local development environments. It operates as an MCP server, an open standard that allows AI agents to interact with external data and tools. This connection capability extends to various services, providing secure exposure for local HTTP or TCP services, including databases and gRPC. It also supports auto-detection for webhooks from providers such as Replicate, fal.ai, RunPod, Modal, and OpenAI, integrating them into AI workflows.

Core features and tools

Pipepie exposes several CLI tools: pie setup configures the server, pie login connects clients, and pie connect tunnels HTTP, TCP, or password-protected URLs, and multi-tunnels from a pipepie.yaml file. pie logs inspects webhooks. It includes AI tool presets like pie connect --ollama, pie connect --comfyui, pie connect --n8n, and pie connect --tma. The built-in pie mcp server provides AI tools direct access to tunnels, requests, and pipeline traces.

Who is it for?

Pipepie is primarily for developers and AI agents who need to securely expose local development servers to the internet or require AI pipeline tracing and webhook inspection. It suits those needing AI tools to directly inspect webhooks, replay requests, manage tunnels, and debug pipelines. Confirmed compatible AI clients include Claude, Cursor, Vercel AI SDK, and Ollama. Skip this if you do not require secure tunneling or AI agent interaction with external tools.

How to get started

Access is granted by possessing the correct public key, authenticating the server using the Noise NK protocol. This system relies on end-to-end encryption. Pipepie supports SSE/streaming transport protocol, compatible with Vercel AI SDK and Ollama.

Use cases

• Securely exposing local development servers to the internet for HTTP or TCP services like databases and gRPC.
• Providing AI pipeline tracing and webhook inspection, with auto-detection for webhooks from providers like Replicate, fal.ai, RunPod, Modal, and OpenAI.
• Enabling AI tools to directly inspect webhooks, replay requests, manage tunnels, and debug pipelines through the built-in MCP server.

Pipepie offers a secure method for connecting local environments to external services and AI pipelines. Its focus on end-to-end encryption and the Model Context Protocol makes it a valuable asset for developers and AI agents. The tool provides essential capabilities for debugging and managing AI interactions with webhooks and tunnels. It is a capable solution for integrating local development with advanced AI workflows. Start by using "pie setup" to configure a server and expose a local service.