kentjuno/kjclicontroller

★ 16HTMLAudience · developerComplexity · 3/5Setup · moderate

Mindmap

mindmap
  root((repo))
    Gateway Core
      Rust backend
      Cargo build
      localhost 8080
    AI Routing
      Claude Code
      Gemini CLI
      Codex CLI
    Chat Interface
      Dark mode UI
      At-mention routing
      Browser based
    Multi Agent
      Conductor model
      Shared log file
      Step delegation
    Security
      Token auth
      Config file
      Local only
    Client Support
      OpenAI API format
      Python helper
      Retry handling

mindmap root((repo)) Gateway Core Rust backend Cargo build localhost 8080 AI Routing Claude Code Gemini CLI Codex CLI Chat Interface Dark mode UI At-mention routing Browser based Multi Agent Conductor model Shared log file Step delegation Security Token auth Config file Local only Client Support OpenAI API format Python helper Retry handling

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Switch between Claude, Gemini, and Codex from one browser chat without juggling separate tools or terminals.

USE CASE 2

Let one AI break a big task into steps and hand each step to a different AI model automatically.

USE CASE 3

Connect any app that already talks to OpenAI to this gateway so it can use multiple AI tools without code changes.

USE CASE 4

Keep AI usage private on your local machine, no prompts leave your computer except through the CLI tools you control.

Tech stack

RustCargoHTMLPythonOpenAI API formatWebServer localhost

Getting it running

Difficulty · moderate Time to first run · 30min

Requires Rust and Cargo installed. Claude Code, Gemini CLI, and/or Codex CLI must be set up separately. Set auth token in config file before first run. Windows users get a system tray icon.

No license was mentioned in the explanation.

In plain English

KJCLIController is a local network gateway built in Rust. It runs on your computer and lets you use three different AI command-line tools (Claude Code from Anthropic, Gemini from Google, and Codex from OpenAI) through a single shared web address. Instead of switching between separate tools, you send your request to one place and the gateway forwards it to whichever AI you pick. The gateway speaks a standard API format originally created by OpenAI, so any application that already knows how to talk to OpenAI can talk to this gateway without code changes. It includes a dark-mode chat interface you open in a browser at localhost:8080/chat. From that interface you can type @claude, @gemini, or @codex to route a message to a specific model. On Windows, the application hides itself in the system tray after startup so it stays out of the way. A newer feature called multi-agent orchestration lets one AI act as a conductor that breaks a task into steps and delegates each step to a different AI. All the agents share a running log file so they can see what the others have already done. The gateway exposes a special endpoint that any AI agent can read to learn how to participate in this pattern, and it also serves a small Python helper script that handles retries and error handling automatically. Security is handled through a token you set in a config file. Any request to the gateway must include that token in its header, which keeps other devices on your local network from using your AI quota. The gateway runs everything as subprocesses on your own machine, so no prompts are sent to any external server that you do not already control directly through the CLI tools. The README covers setup, available endpoints, a Python client example, and the orchestration protocol in detail. The project is written in Rust and builds with the standard Cargo toolchain.

Copy-paste prompts

Prompt 1

I have KJCLIController running at localhost:8080. Walk me through sending a request to the /chat endpoint using Python's requests library, routing it to @claude, and handling the response.

Prompt 2

Using KJCLIController's multi-agent orchestration feature, help me design a workflow where one AI acts as a conductor to break a coding task into steps and delegates each step to Claude, Gemini, or Codex.

Prompt 3

I want to connect my existing OpenAI-compatible app to KJCLIController instead of OpenAI directly. What base URL and auth header changes do I need to make?

Prompt 4

Show me a minimal Python script that uses the KJCLIController Python helper to send a prompt with automatic retries and error handling.

Prompt 5

How do I set up the token in the KJCLIController config file and pass it in request headers so other devices on my local network cannot use my AI quota?

Open on GitHub → Explain another repo

← kentjuno on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.