Context Pipe

Name: Context Pipe
Author: luismichio

@luismichio

Repository

About Context Pipe

No overview available yet

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "context-pipe": {
      "command": "context-pipe-server",
      "args": [],
      "env": {
        "PIPE_CONFIG_PATH": "pipes.json"
      }
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is Context Pipe?

Context Pipe is a high-performance orchestration layer that chains AI tool calls through context streams, inspired by Unix terminal piping. It integrates MCP tools, shell commands, scripts, and binaries into named pipes that refine raw data into dense, high-signal content before it reaches an LLM. It is built for developers and AI engineers who need to reduce context window noise and make tool outputs composable and production-grade.

How to use Context Pipe?

Install the Python package (mcp-context-pipe) via pip, then define pipes in a pipes.json file (project-level) with optional global overrides in ~/.mcp-pipe.json. Use the mcp-pipe CLI to run named pipes (mcp-pipe run <pipe>), invoke MCP tools directly from the shell (mcp-pipe tool <server> <tool>), or inject IDE hooks via pipe_onboard. The Rust-native cpipe binary is also available as a standalone CLI or Tauri sidecar for sub‑2ms startup latency.

Key features of Context Pipe

Unix pipe model for AI: chain any stdin/stdout tool into a named pipe
MCP node type: call any MCP tool as a first-class pipe node
Dynamic pipes: agents construct ad‑hoc node lists at runtime
A2A agent handoff: distil Agent A’s output before Agent B sees it
Context Balance Sheet: per‑run accounting of input/output bytes and latency
Shadow MCP Registry: keep utility MCP servers invisible until needed

Use cases of Context Pipe

Crawl a webpage, filter for security keywords, convert to Markdown, and distill to a summary
Pass an error log through a semantic sifter, then pipe the output into a grep for critical issues
Hand off processed context between agents in CrewAI, LangGraph, or Google ADK without bloating the context window
Automatically refine tool outputs in an IDE (Cursor, VS Code, Windsurf, etc.) via subconscious interceptors
Audit and trace the entire context supply chain across multiple refineries in one pipe

FAQ from Context Pipe

What dependencies does Context Pipe require?

Python 3.10+ is required for the MCP server and Python orchestrator. The Rust native core (cpipe) is optional but recommended for sub‑2ms startup latency; it is available as a Tauri sidecar, standalone CLI, or Cargo library.

How is Context Pipe configured?

Pipe definitions live in pipes.json (project‑level) and optionally ~/.mcp-pipe.json (global, merged with local precedence). No code changes are needed to add, modify, or re‑route pipes.

What transport does Context Pipe use?

Context Pipe nodes are real OS subprocesses communicating via stdin/stdout byte streams. The orchestration engine uses shell=False (no injection surface) and enforces per‑node timeouts. IDE hooks use the MCP protocol over stdio.

Does Context Pipe support authentication?

Authentication is not covered in the readme; pipes trust the local environment. MCP tools called via mcp-pipe tool inherit the servers’ own authentication (e.g., GitHub tokens) from their server registration in pipes.json.

What are the known limitations of Context Pipe?

For the OpenCode IDE, the AGENTS.md SOP mandate is the active strategy (no native hook yet). The orchestrator enforces non‑UTF8 robustness with errors="replace" and null‑safe reading. All nodes run with shell=False to eliminate injection risks.