ncp

Overview

NCP (Natural Context Provider) solves the problem of cognitive overload when AI assistants have access to too many MCP tools. Instead of exposing dozens of individual tools, NCP provides a unified interface with just three simple tools: 'find' for discovering capabilities, 'code' for executing TypeScript, and 'run' for individual tool execution. This reduces token usage by up to 97% and improves response times by 5x. NCP also supports project-level configuration, scheduling, and custom Photons (lightweight TypeScript MCPs without npm publishing).

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• find

  Search for any tool, skill, or Photon using natural language

  Try:I need to read a fileTry:help me send an emailTry:search for something online
• code

  Execute TypeScript directly with access to all available MCPs

  Try:await github.create_issue({title: 'New feature', body: 'Details here'})Try:write a function to process this data
• run

  Execute tools individually when code mode is disabled

  Try:run filesystem read_file --path '/tmp/test.txt'Try:run web search --query 'latest MCP news'

Comparable tools

Installation

npm install -g @portel/ncp

{
  "mcpServers": {
    "ncp": {
      "command": "ncp"
    }
  }
}

FAQ

What is the difference between NCP and regular MCPs?

Regular MCPs expose dozens of individual tools that overwhelm AI assistants. NCP unifies these tools into just three simple tools (find, code, run) while managing all the complexity behind the scenes.

How does NCP reduce token usage?

Instead of loading dozens of tool schemas (100k+ tokens), NCP exposes just three simple tools (2.5k tokens), reducing token usage by up to 97% and allowing AI to focus on actual work.

Compare ncp with