Ixel MAT

Multi-Agent Terminal by IxelAI. Run multiple AI providers side-by-side from the terminal, compare answers in real time, and synthesize a faster consensus when needed.

Install

Linux / macOS

curl -fsSL https://raw.githubusercontent.com/OpenIxelAI/ixel-mat/main/install.sh | bash

Windows PowerShell

powershell -NoProfile -ExecutionPolicy Bypass -Command "irm https://raw.githubusercontent.com/OpenIxelAI/ixel-mat/main/install.ps1 | iex"

After install:

ixel help
ixel status
ixel setup
ixel

Why Ixel MAT?

• Real-time comparison: /full streams each model as soon as it responds.
• Faster synthesis: /consensus starts once enough strong answers arrive instead of waiting for every provider.
• Operator-first UX: local config, local secrets, health/status commands, and terminal-native workflows.

Demo

▸ Explain zero trust architecture
3/5 agents responded

╭──────────────────────────── OpenAI ────────────────────────────╮
│ Zero trust assumes no user, device, or network path should be │
│ trusted by default. Every request must be authenticated,       │
│ authorized, and continuously evaluated.                        │
╰────────────────────────────────────────────────────────────────╯
╭──────────────────────────── Claude ────────────────────────────╮
│ Zero trust replaces perimeter trust with identity, policy,    │
│ and verification at every step of access.                     │
╰────────────────────────────────────────────────────────────────╯
╭──────────────────────────── Gemini ────────────────────────────╮
│ processing... 6.4s                                             │
╰────────────────────────────────────────────────────────────────╯

Supported Providers

Provider	Auth method	Models
OpenClaw Gateway	IXELMAT_GATEWAY_TOKEN	Any model exposed by your OpenClaw gateway
OpenAI	OPENAI_API_KEY	gpt-4o, gpt-4o-mini, gpt-4-turbo, o1, o3, gpt-5.4
Anthropic	ANTHROPIC_API_KEY	claude-opus-4, claude-sonnet-4-5, claude-haiku-3
xAI	XAI_API_KEY	grok-4, grok-3, grok-3-mini
Google Gemini	GOOGLE_API_KEY	gemini-2.5-pro, gemini-2.5-flash, gemini-2.0-flash

Core Commands

Command	What it does
ixel	Launch the interactive MAT terminal
ixel status	Show provider health, agent connectivity, secrets, and warnings
ixel setup	Configure providers and write config + .env
ixel agents	Probe configured agents
ixel models	Show providers, models, and auth state
pytest tests/	Run the automated test suite

How it works

• mat.py runs the interactive terminal UI
• cli.py handles setup, status, models, doctor, and config commands
• agents/ provides HTTP, WebSocket, subprocess, and one-shot adapters
• modes/full.py and modes/consensus.py drive compare/synthesis workflows
• config/ handles TOML loading, validation, setup, and secrets

Quick Start from Source

git clone https://github.com/OpenIxelAI/ixel-mat.git
cd ixel-mat
python3 -m pip install -r requirements.txt
python3 cli.py help
python3 cli.py status
python3 cli.py setup
python3 mat.py

Run tests:

pytest tests/

Config + Secrets

• Config: ~/.config/ixel-mat/config.toml
• Secrets: ~/.config/ixel-mat/.env
• Secrets file permissions are automatically set to 600
• Starter config: config.example.toml

Roadmap

• Token-level provider streaming
• Single-agent interactive mode (/agent <name>)
• pyproject.toml + pipx packaging
• Better local subprocess workflows for CLI-native agents
• Compare/diff mode for disagreements across models