quarkus-chat-ui

quarkus-chat-ui is a multi-provider web UI for Large Language Models. It wraps Claude Code CLI, OpenAI Codex CLI, or any OpenAI-compatible HTTP server (vLLM, Ollama) and streams responses to your browser in real time.

Multiple instances can talk to each other via MCP — enabling autonomous agent-to-agent conversations while humans watch from their browsers.

Features

Multiple LLM providers — Claude Code CLI, OpenAI Codex CLI, vLLM, Ollama, any OpenAI-compatible API
Streaming responses — Server-Sent Events (SSE) for real-time token streaming
Prompt queue — Queue multiple prompts; they execute automatically in order
MCP server — Each instance exposes itself at /mcp for agent-to-agent communication
Multi-agent conversation — Instances can call each other; replies flow back automatically
Tool execution — Bash, Read, Write, Edit, Glob, Grep, WebSearch, WebFetch (CLI providers)
Interactive permission dialogs — Tool permission requests forwarded to the Web UI
Theme support — 10 built-in themes (5 dark + 5 light)
Slash commands — Provider-specific commands (/model, /compact, /clear, …)
Keyboard modes — Default, Mac, and Vim keybindings
URL fetch — Fetch and extract text from URLs for inclusion in prompts

Providers

`chat-ui.provider`	Backend	Auth
`claude`	Claude Code CLI	`ANTHROPIC_API_KEY`
`codex`	OpenAI Codex CLI	`OPENAI_API_KEY`
`openai-compat`	Any OpenAI-compatible HTTP server (vLLM, Ollama, …)	optional API key

Quick Start

Native Executable (no JDK required)

Pre-built native executables are available on the Releases page. No JVM, no JDK — just download and run.

Platform	Binary
Linux x86_64	`quarkus-chat-ui-linux-amd64`
Linux ARM64	`quarkus-chat-ui-linux-arm64`
macOS Intel	`quarkus-chat-ui-macos-amd64`
macOS Apple Silicon (M1/M2/M3)	`quarkus-chat-ui-macos-arm64`
Windows x64	`quarkus-chat-ui-windows-amd64.exe`

# Linux / macOS
./quarkus-chat-ui-linux-amd64 -Dchat-ui.provider=claude -Dquarkus.http.port=28010

# Windows PowerShell
.\quarkus-chat-ui-windows-amd64.exe -Dchat-ui.provider=claude -Dquarkus.http.port=28010

Build from Source

Prerequisites: JDK 21+ and Maven 3.x

git clone https://github.com/scivicslab/quarkus-chat-ui
cd quarkus-chat-ui
mvn install -DskipTests

Run with Claude Code CLI:

java -Dchat-ui.provider=claude \
     -Dquarkus.http.port=28010 \
     -jar app/target/quarkus-app/quarkus-run.jar

Run with vLLM or Ollama:

java -Dchat-ui.provider=openai-compat \
     -Dchat-ui.servers=http://localhost:11434/v1 \
     -Dquarkus.http.port=28010 \
     -jar app/target/quarkus-app/quarkus-run.jar

Open http://localhost:28010 in a browser.

Configuration

All properties can be passed as -D flags or set in application.properties.

Property	Default	Description
`chat-ui.provider`	`claude`	LLM provider: `claude`, `codex`, or `openai-compat`
`chat-ui.servers`	`http://localhost:8000`	Server URLs for `openai-compat` (comma-separated)
`chat-ui.default-model`	(provider default)	Model name override
`chat-ui.api-key`	(env var)	API key (prefer env vars)
`chat-ui.title`	`Coder Agent`	Browser tab and header title
`chat-ui.keybind`	`default`	Keyboard mode: `default`, `mac`, or `vim`
`chat-ui.gateway-url`	(none)	MCP Gateway URL for multi-agent routing
`quarkus.http.port`	`8090`	HTTP listen port

MCP Server

Each instance exposes itself as an HTTP MCP server at /mcp. Available tools:

Tool	Description
`submitPrompt`	Send a prompt to the LLM (queued, async). Accepts `_caller` for agent-to-agent replies.
`getPromptStatus`	Check if the LLM is still processing
`getPromptResult`	Retrieve the completed response
`cancelRequest`	Interrupt the current LLM request
`getStatus`	Current model, session ID, and busy state
`listModels`	Available model names

claude mcp add --transport http chat-ui-28010 http://localhost:28010/mcp

curl -X POST http://localhost:8888/api/servers \
  -H 'Content-Type: application/json' \
  -d '{"name": "chat-ui-28010", "url": "http://localhost:28010", "description": "Chat UI on port 28010"}'

Multi-Agent Conversation

When Instance B receives a prompt with _caller pointing back to Instance A, it enriches the prompt with position awareness and reply instructions:

[Context]
You are running on: http://localhost:28020
Received via MCP from: localhost:28010

[Message]
What should we work on today?

[How to Reply]
Use callMcpServer tool:
- serverUrl: http://localhost:28010
- toolName: submitPrompt
- arguments: {"prompt": "your reply", "_caller": "http://localhost:28020"}

The LLM reads this, formulates a reply, and calls submitPrompt on Instance A. The conversation continues autonomously while humans watch from their browsers.

Architecture

Built on POJO-actor, a lightweight actor-model library for Java 21. Each concern — chat session, side questions, queue management, stall detection — runs in its own actor. Blocking I/O runs on virtual threads. There are no synchronized blocks in the application code.

See the blog post: quarkus-chat-ui: A Web Front-End for LLMs, and a Real-World Case for POJO-actor

Multi-Agent Setup Tutorials

Step-by-step guides for setting up multi-agent communication:

Tutorial	Description
Two Claude Code CLI Agents	Two instances both using Claude Code CLI
Claude Code CLI + Codex	Heterogeneous setup with Anthropic and OpenAI
Claude Code CLI + Local LLM	Using claw-code-local with Ollama

For three or more agents, see MCP Gateway for name-based routing.

Features​

Providers​

Quick Start​

Native Executable (no JDK required)​

Build from Source​

Configuration​

MCP Server​

Multi-Agent Conversation​

Architecture​

Multi-Agent Setup Tutorials​