Connect

Add Pauhu to your client in two minutes.

Pauhu Fusion is a remote MCP server at mcp.pauhu.dev. Add it with the API key issued at checkout. No install, no third-party service.

EasiestPaste this prompt into your AI chat

In an agentic client (Claude Code, Cursor, Copilot agent), paste this and your assistant adds the server for you.

Add a remote MCP server named "pauhu" at https://mcp.pauhu.dev/mcp using HTTP (streamable) transport, with the header "Authorization: Bearer YOUR_KEY" (replace YOUR_KEY with the key from my Pauhu checkout). Then list the pauhu tools to confirm it connected.

Or set it up by hand, the way each provider documents it:

claude mcp add --transport http pauhu https://mcp.pauhu.dev/mcp \
  --header "Authorization: Bearer <YOUR_KEY>"

Official Claude Code MCP docs

// .vscode/mcp.json
{ "servers": { "pauhu": { "type": "http", "url": "https://mcp.pauhu.dev/mcp",
  "headers": { "Authorization": "Bearer ${input:pauhu-key}" } } } }

Official VS Code Copilot MCP docs

// .cursor/mcp.json
{ "mcpServers": { "pauhu": { "url": "https://mcp.pauhu.dev/mcp",
  "headers": { "Authorization": "Bearer ${env:PAUHU_KEY}" } } } }

Official Cursor MCP docs

# ~/.codex/config.toml
[mcp_servers.pauhu]
url = "https://mcp.pauhu.dev/mcp"
bearer_token_env_var = "PAUHU_MCP_KEY"

Official Codex MCP docs

And any other client that speaks MCP.

  • Claude Code
  • Claude Desktop
  • ChatGPT
  • VS Code Copilot
  • Cursor
  • Codex
  • Grok
  • Windsurf
  • Cline
  • Continue
  • Zed
  • JetBrains
  • Gemini CLI
  • Mistral
Safe by design

What it does for you, and where it stops.

What it does

  • Builds the operation from published standards and documented APIs.
  • Returns the same code for the same request, every time.
  • Never invents an API. Every line traces back to a real source.
  • Reports the tokens and energy each call avoided, on a receipt.

What it does not

  • Is not a general chatbot over your code folder.
  • Does not replace a senior engineer.
  • Does not take your repository; only the part a request needs crosses the wire.
  • Does not run code or take actions on your machine.
Use cases

Where it earns its place.

Stdlib and standard ops

File, string, integer math, collections, json, http, hashing, built from published sources instead of guessed.

Right-API code

Built against published standards and library APIs, so the call uses the documented signature, not a plausible one.

Bounded AI context

Send your model only the context a task needs. Less token load, less data shared.

Reserve access