OmniRoute MCP Client: CLI-Driven AI Gateway with Remote Control
OmniRoute MCP Client: CLI-Driven AI Gateway with Remote Control
OmniRoute is a TypeScript-based AI gateway and MCP client designed to streamline access to a vast ecosystem of large language models. It provides a single endpoint to over 290 providers (including 90+ free options) and 500+ models, encompassing Kimi, Claude, GPT, OpenAI, Gemini, GLM, DeepSeek, and MiniMax. What sets OmniRoute apart is its comprehensive command-line interface (CLI), which isn't just for initial setup but serves as a full cockpit for managing the gateway, whether it's running locally or on a remote server.
Full CLI Control: Local and Remote
OmniRoute's CLI offers over 80 commands, moving far beyond basic start functionality. This deep command surface allows developers to interact with the gateway, configure providers, manage models, and even control an OmniRoute instance running on a remote server, all from their local machine.
For local operations, essential commands include:
omniroute # serve gateway + dashboard (port 20128)
omniroute chat # interactive TUI chat client (slash: /model /combo /skill /memory)
omniroute setup # guided first-run wizard
omniroute doctor # diagnose providers, ports, native depsThe true power of the CLI emerges with its remote mode. You can deploy OmniRoute on a VPS and drive it from your laptop using the same CLI. After a single login with a scoped access token, every subsequent command targets the remote instance. This enables seamless management of your AI gateway infrastructure from anywhere.
omniroute connect 192.168.0.15 # password → scoped token, saved as a context
omniroute models list # ← runs against the REMOTE server
omniroute configure codex # ← picks a remote model, writes a local Codex profile
omniroute tokens create --name ci --scope read # mint narrower tokens for other machines
omniroute contexts use default # ← switch back to the local serverAccess tokens are granular, supporting read, write, and admin scopes, with process-spawning routes restricted to loopback-only for security.
Agent-Driven Automation via MCP/A2A
Beyond direct human interaction, OmniRoute is built with open agent protocols (MCP/A2A) that allow an AI agent to drive it autonomously. This means an agent can issue commands, configure settings, and manage model interactions directly through the CLI, opening up possibilities for self-optimizing AI workflows, automated testing, or dynamic resource allocation based on real-time needs. The CLI's extensive command surface, covering areas like providers, keys, models, cache, compression, cost, usage, quota, health, and resilience, makes it a rich target for agent-based control.
Zero-Config Start and Advanced Compression
OmniRoute is designed for immediate use. A fresh installation requires no API keys or configuration to get started. After npm i -g omniroute, the server boots on localhost:20128. You can then point any OpenAI-compatible tool (like Claude Code, Cursor, or Cline) to http://localhost:20128/v1. The auto model combo, pre-wired with keyless free providers like OpenCode Free and Felo, responds instantly out of the box.
For optimizing token usage and cost, OmniRoute includes a sophisticated 7-option compression pipeline (off, lite, standard, aggressive, ultra, RTK, stacked). RTK+Caveman compression can save between 15-95% of tokens. This system includes features like command-output compression, filters, trust/verify mechanisms, and raw-output recovery. Compression rules are defined via JSON rule-pack schemas for Caveman and RTK filters, supporting language detection and custom rule-pack authoring.