Connects to a local Ollama instance (default http://localhost:11434) to run local models in parallel with cloud-based AI providers.
Integrates with OpenAI's API and ChatGPT subscriptions to run OpenAI models in parallel, using a local Codex login session or the OPENAI_API_KEY environment variable.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@MegaPad How do I optimize this SQL query?"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Scientific Multi-Model Benchmarking & Token Cost Slasher for Claude Code, Cursor, and Terminal.
Quickstart · Token & Cost Analytics · Claude Code / Cursor · Piping Support
Run a multi-model race on any coding question in one command:
# Global / Local
pad "How do I optimize this query?"
# Or via npx (zero install)
npx megapad "How do I optimize this query?"Executes across your available models in parallel and displays latency, generation speed (tok/s), exact token metrics, and cost savings vs Claude.
Related MCP server: Zen MCP Server
Add MegaPad into your AI coding agent as a Native Model Context Protocol (MCP) server:
pad installAuto-detects and writes config for Claude Code (~/.claude.json), Cursor (~/.cursor/mcp.json), and Codex CLI (~/.codex/config.toml).
Use the /pad slash command directly:
/pad "How do I optimize this query?" — Instant multi-model race & telemetry
/pad review — Multi-model code review council
/pad status — Inspect connected accounts & model subscriptions
Every query gives you high-resolution latency, tokens per second, and dollar cost comparisons:
Provider / Model | Status | Latency | Speed | Cost (USD) | vs Claude Cost |
DeepSeek V4.1 Flash | ✔ OK | 890ms | 185 tok/s | $0.00010 | ⚡ 98% Cheaper |
Gemini 3.8 Flash | ✔ OK | 940ms | 145 tok/s | $0.00000 | 🎉 100% Free ($0) |
ChatGPT (Subscription) | ✔ OK | 820ms | 125 tok/s | $0.00000 | 🎉 $0 Extra (Sub) |
GPT-6 Astra / Sol | ✔ OK | 1,020ms | 130 tok/s | $0.00280 | ⚡ 30% Cheaper |
Claude Fable 5.1 / Sonnet 5 | ✔ OK | 1,140ms | 110 tok/s | $0.00400 | Base (100%) |
Pipe terminal output, compiler errors, or git diffs directly into MegaPad:
# Diagnose error logs across multiple models
cat error.log | npx megapad "Diagnose this failure"
# Multi-model review of your git staging
git diff --staged | npx megapad "Review these changes for race conditions"MegaPad automatically detects your existing CLI subscriptions and local environment keys:
ChatGPT / OpenAI: Reads your local Codex login session or OPENAI_API_KEY.
Claude: Uses your active Claude Code session or ANTHROPIC_API_KEY.
DeepSeek: Reads DEEPSEEK_API_KEY (from environment or ~/.env).
Gemini: Free tier quota (15 RPM free forever) or GEMINI_API_KEY.
Local Models: Connects to local Ollama / LM Studio on http://localhost:11434.
MIT © Eugene Finch & MegaPad Contributors
This server cannot be deployed
Multi-LLM council: 25+ frontier models in parallel, consensus scoring, verdict-first code review.
One identity across Claude Code, Codex, Cursor, Gemini, Windsurf: shared inbox and handoffs.
AI routing, memory, guardrails, and governance. Routes across Claude, GPT, Gemini.
Build, test, monitor & improve AI with Future AGI via Claude, Cursor, Windsurf & more.
