OpenLadonx

OpenLadonx

Open desktop workspace for modern AI coding agents — Codex-style and Claude Code-style workflows in one native Tauri app.

npm install · npm run tauri:dev · OpenAI Responses / Anthropic Messages

English · 简体中文

OpenLadonx is an open desktop workspace for modern AI coding agents. It brings Codex-style and Claude Code-style workflows into one native Tauri app: chat threads, workspace context, git review, terminals, model routing, skills, plugins, MCP server status, and local desktop integrations.

Codex and Claude Code both offer powerful agent experiences, but their official desktop clients are closed-source. OpenLadonx takes a different path: it integrates the desktop-side workflow patterns developers already like, opens most of the implementation, and gives builders a real codebase they can study, modify, fork, and reshape into the agent desktop they actually want.

Use the built-in provider path, or connect your own OpenAI Responses-compatible and Anthropic Messages-compatible model endpoints. OpenLadonx is designed for developers, teams, and agent tinkerers who want a local-first command center instead of another locked-down black box or browser tab.

What OpenLadonx Does

OpenLadonx turns the desktop client itself into something developers can own. Instead of treating the agent UI as a sealed product, it exposes the practical building blocks behind a serious AI coding workspace: model connections, prompt and instruction surfaces, local workspace awareness, terminal access, git context, plugin state, MCP visibility, and native app integration.

• Unifies agent workflows for Codex-style and Claude Code-style development in a single desktop app.
• Opens the desktop layer so developers can inspect, fork, and rebuild the parts of the agent experience that are usually hidden inside closed-source clients.
• Connects your own models through OpenAI Responses-compatible /v1/responses endpoints or Anthropic Messages-compatible /v1/messages endpoints.
• Keeps real work close with conversations, workspaces, files, terminals, diffs, branches, issues, and pull requests in one native surface.
• Makes extension first-class with skills, plugins, prompts, MCP tool-call rendering, and workspace AGENTS.md editing.
• Gives local configuration control for model credentials, Codex config.toml, plugin state, MCP status, and desktop settings.

If you are building an internal agent platform, experimenting with custom models, designing a new coding-agent UX, or just tired of waiting for closed desktop clients to expose the one feature you need, OpenLadonx is meant to be a strong starting point.

Model Routing

OpenLadonx supports two custom API protocol families from Settings:

To add a model:

1. Open Settings.
2. Go to the model/API key section.
3. Click Add model.
4. Choose OpenAI/Response or Anthropic/Messages.
5. Enter the base URL, API key, and one or more model IDs.
6. Run the built-in test request.
7. Save the configuration. The models will appear in the model selector.

Custom model settings are stored locally. Do not commit API keys, local auth files, generated config, or machine-specific signing settings.

Quick Start

Install dependencies:

npm install

npm install also runs local asset/resource sync. The Tauri resources under src-tauri/resources/ are intentionally ignored because they include local tool binaries. If they are missing, run:

npm run sync:tauri-resources

Run the frontend only:

npm run dev

Launch the full Tauri desktop app:

npm run tauri:dev

Run the strict environment check before desktop builds:

npm run doctor:strict

Development Commands

npm run build
npm run sync:tauri-resources
npm run sync:tauri-resources:all
npm run lint
npm run test
npm run typecheck
cargo check --manifest-path src-tauri/Cargo.toml

sync:tauri-resources prepares the current platform binaries and generated resource templates before local dev. sync:tauri-resources:all validates all macOS/Windows resource slots and is used by the desktop release script.

For focused Rust/Tauri validation:

cd src-tauri
cargo check
cargo test

Feature Surface

• Workspace-aware chat threads with file references and token-aware autocomplete.
• File previews, terminal panes, branch helpers, git diffs, and PR-oriented review context.
• Model/API key management for compatible Responses and Messages endpoints.
• Skills, plugins, prompt management, and MCP server status surfaces.
• Local editors for global and workspace-level AGENTS.md instructions.
• Codex config.toml editing from the desktop UI.
• Multi-language UI assets and English/Chinese README documentation.

Project Layout

.
├── public/              # Static web assets
├── src/                 # React application code
│   ├── features/        # Feature modules
│   ├── hooks/           # Shared React hooks
│   ├── services/        # App and Tauri service clients
│   ├── styles/          # Global and design-system styles
│   ├── test/            # Vitest setup
│   └── utils/           # Shared helpers
├── src-tauri/           # Tauri configuration and Rust code
│   ├── resources/       # Ignored generated templates and local tool binaries
│   ├── src/             # Rust commands, services, and app wiring
│   └── tests/           # Rust integration tests
├── scripts/             # Build, doctor, release, and maintenance scripts
└── docs/                # Project documentation and static docs site

Prerequisites

• Node.js and npm.
• Rust toolchain with Cargo.
• Tauri system dependencies for your platform.
• A model credential from a supported default provider, or a compatible custom endpoint.

Configuration

OpenLadonx reads and writes local desktop settings, Codex configuration, and workspace instruction files. Typical local configuration includes:

• API keys and base URLs.
• Custom Responses API and Messages API model lists.
• Global and workspace AGENTS.md instructions.
• Codex config.toml content.
• Plugin enablement and MCP configuration managed by the underlying agent environment.

For Anthropic-compatible providers, add a model from Settings with the Anthropic/Messages protocol. Use a base URL that exposes a Messages-compatible /v1/messages endpoint, enter the provider API key, and list the model IDs you want to show in the selector, such as Claude-family models or compatible router model names. OpenLadonx keeps this configuration local so you can switch between official Anthropic access, private gateways, and third-party compatible providers without changing the app code.

The repository includes .testflight.local.env.example as a reference for local release configuration. Copy it to an ignored local file when needed.

Contributing

• Use 2-space indentation for TypeScript and TSX.
• Name React components in PascalCase, hooks as useSomething, utilities in camelCase, and Rust functions in snake_case.
• Prefer configured aliases such as @/, @app/, @services/, and @utils/.
• Keep frontend IPC contracts aligned with Rust Tauri commands and payload types.
• Add focused tests near the feature or module you change.
• Run npm run typecheck, npm run lint, and relevant tests before opening a pull request.

License

This project is licensed under the terms in LICENSE.

Contact

• Email: greentm372@gmail.com
• Backup email: 1691587386@qq.com
• WeChat: