Sandbox
@SakuraByteCore/codexmate

Local dashboard for coding agents

Codex Mate gives you one local place to manage providers, sessions, skills, and tasks across several coding agents. It works through a CLI, a web UI, and an MCP server, while storing config and session data in your home directory.

344 stars35 forksJavaScriptUpdated 6d ago
Who it's for

Builders who want one local control plane for multiple coding agents and their sessions.

What it delivers

You can switch between agents, inspect past work, and orchestrate tasks without juggling separate configs.

What it does

Unified session browser

Search, inspect, filter, convert, and export sessions from Codex, Claude Code, Gemini CLI, CodeBuddy Code, Pi, and others.

Provider management

Switch and configure providers and models for Codex, Claude Code, OpenCode, KiloCode, OpenClaw, and Pi from the CLI or web UI.

Local agent bridges

Normalize Codex requests for OpenAI-compatible UIs and connect Claude Code to OpenAI Chat Completions-compatible providers and Ollama.

Skills marketplace

Import and share skills between different agent apps from a local-first skills market.

Prompt file editor

Edit global and project-level `CLAUDE.md` and `AGENTS.md` with shared presets and auto-detected project paths.

Task orchestrator

Plan and execute tasks with dependency tracking and logs.

MCP integration

Expose local tools and resources through MCP stdio.

Health checks and cleanup

Probe provider routes, surface broken configs, and remove failed entries without touching healthy ones.

How to get it

  1. 1Run
    brew tap SakuraByteCore/codexmate
    brew install codexmate
  2. 2Run
    npm install -g codexmate
    codexmate run
  3. 3If the default Web UI port 3737 is unavailable, Codex Mate automatically tries the next…
    CODEXMATE_PORT=8080 codexmate run
  4. 4Windows PowerShell
    $env:CODEXMATE_PORT=8080; codexmate run
  5. 5Run
    curl -fsSL https://raw.githubusercontent.com/SakuraByteCore/codexmate/main/scripts/install.sh | bash

README

Codex Mate logo

Codex Mate

One dashboard for all your local AI coding agents. Switch providers, manage sessions, and orchestrate tasks across Codex, Claude Code, Gemini CLI, CodeBuddy Code, Pi, OpenCode, KiloCode, and OpenClaw. Zero cloud, local-first control plane.

[Documentation] [Quick Start] [简体中文] [Tiếng Việt]

Version Build Downloads Install Platform Node License Stars Issues On StackMap


Codex Mate Codex provider configuration

Codex Mate OpenCode provider configuration

Codex Mate unified session browser


[!TIP] Local First: All configurations and sessions are stored in your home directory. No telemetry, no cloud accounts required.

[!IMPORTANT] This project is currently in early stage. We are seeking developers to help build the local agent ecosystem!

What is Codex Mate?

Have you ever felt overwhelmed by managing multiple local AI agents? Each has its own config format, session storage, and skills directory.

Codex Mate offers a unified control plane to bring order to the chaos. It's a local-first CLI + Web UI designed to manage Codex, Claude Code, OpenCode, KiloCode, OpenClaw, and Pi seamlessly.

What's So Special?

Unlike simple wrappers, Codex Mate acts as a Local Agent Bridge:

  • Unified Session Browser: Search, inspect, filter, and export local sessions across Codex, Claude Code, Gemini CLI, CodeBuddy Code, and Pi from one place.
  • OpenAI-Compatible Bridge: Use Codex with any OpenAI-compatible UI by normalizing the Responses API; the built-in Codex conversion also fills and normalizes Codex fingerprint headers such as User-Agent, Version, OpenAI-Beta, and Originator so upstream providers see an official Codex CLI-shaped request.
  • Claude Provider Bridge: Connect Claude Code to OpenAI Chat Completions-compatible providers and Ollama through the built-in local Claude-compatible proxy.
  • OpenCode Provider Control: Manage OpenCode provider/model selection with a CodexMate-owned provider store under ~/.codexmate, projecting only the active provider into native OpenCode config to avoid polluting or deleting user-owned settings.
  • KiloCode Config Bridge: Configure KiloCode providers from the CLI or Web UI, writing to ~/.config/kilo/kilo.jsonc while preserving stored API keys.
  • Pi Provider Config: Manage Pi providers from the Web UI with lossless round-trip writes to ~/.pi/agent/models.json, auto-fetched remote model lists, and one-click default provider/model switching via ~/.pi/agent/settings.json.
  • Provider Health Cleanup: Probe local Codex and Claude provider routes, surface failed configs in one modal, and bulk-clean selected broken providers without touching healthy or protected entries.
  • Skills Marketplace: A local-first market to share and import skills between different agent apps.
  • Prompt File Editor: Unified editor for global and project-level CLAUDE.md and AGENTS.md with auto-detection of project paths, plus a shared preset pool for reusable prompts.
  • Task Orchestrator: Plan and execute complex tasks with dependency tracking.

Current Progress

FeatureStatusDescription
Provider ManagementSwitch providers/models for Codex, Claude, OpenCode, KiloCode, OpenClaw, and Pi
Live Agent SyncReal-time monitoring of Codex/Claude config & status
Session BrowserSearch, preview, filter, and export sessions across Codex, Claude Code, Gemini CLI, CodeBuddy Code, and Pi
Usage AnalyticsVisualize message trends and top projects
Local Skills MarketCross-app import/export of agent skills
Task QueueDAG-based task execution and logs
OpenAI BridgeConvert Codex Responses API to standard OpenAI format and attach/normalize Codex fingerprints in the built-in conversion
Claude Provider BridgeConnect Claude Code to OpenAI Chat Completions-compatible providers and Ollama via the built-in Claude-compatible proxy
OpenCode Provider StoreKeep multiple OpenCode providers in ~/.codexmate while projecting only the selected provider to native OpenCode config
KiloCode Config BridgeConfigure KiloCode through codexmate kilo config or the Web UI, writing provider URL/model/API key data to ~/.config/kilo/kilo.jsonc
Pi Provider ConfigManage Pi providers in the Web UI writing to ~/.pi/agent/models.json (unknown fields preserved) and switch the default provider/model in ~/.pi/agent/settings.json
Provider Health CheckProbe local Codex/Claude provider routes, highlight failed configs, and bulk-remove selected broken providers safely
Prompt TemplatesReusable prompt plugins with variables
Prompt File EditorEdit global and project-level CLAUDE.md / AGENTS.md with auto-detect, path switching, and a shared preset pool. Applying a preset only updates the editor; save manually to write the file.
MCP IntegrationExpose local tools and resources via MCP stdio
Auto UpdateQuick update CLI via codexmate update

Quick Start

Install via Homebrew (macOS / Linux)

brew tap SakuraByteCore/codexmate
brew install codexmate

Requires Node.js (brew install node if not present).

Install via npm

npm install -g codexmate
codexmate run

If the default Web UI port 3737 is unavailable, Codex Mate automatically tries the next ports (3738, 3739, ...). To force a fixed port, set CODEXMATE_PORT:

CODEXMATE_PORT=8080 codexmate run

Windows PowerShell:

$env:CODEXMATE_PORT=8080; codexmate run

Install via curl (Standalone)

curl -fsSL https://raw.githubusercontent.com/SakuraByteCore/codexmate/main/scripts/install.sh | bash

Supported Agents

  • Codex: npm install -g @openai/codex
  • Claude Code: npm install -g @anthropic-ai/claude-code
  • Gemini CLI: npm install -g @google/gemini-cli
  • CodeBuddy: npm install -g @tencent-ai/codebuddy-code
  • KiloCode: npm install -g @kilocode/cli (kilo / kilocode)
  • OpenCode: install from the official OpenCode docs
  • OpenClaw: see the OpenClaw repository
  • Pi: sessions are discovered from ~/.pi/agent/sessions; provider config lives in ~/.pi/agent/models.json and the default provider/model in ~/.pi/agent/settings.json

Architecture

%%{ init: { 'flowchart': { 'curve': 'catmullRom' } } }%%
flowchart TD
    User([User])
    CLI[CLI]
    WebUI[Web UI]
    MCP[MCP Server]

    subgraph Mate [Codex Mate Core]
        API[HTTP API]
        Config[Config Engine]
        Session[Session Manager]
        Skills[Skills Market]
        Tasks[Task Runner]
    end

    subgraph Local [Local Filesystem]
        CodexDir[~/.codex]
        ClaudeDir[~/.claude]
        ClawDir[~/.openclaw]
        OpenCodeDir[~/.config/opencode]
        MateDir[~/.codexmate]
        State[Sessions/Usage/Trash]
    end

    User --> CLI & WebUI & MCP
    CLI & WebUI & MCP --> API

    API --> Config & Session & Skills & Tasks

    Config --> CodexDir & ClaudeDir & ClawDir & OpenCodeDir & MateDir
    Session --> State
    Skills --> Local

Special Thanks

Special thanks to all contributors for their contributions to Codex Mate ❤️

License

Apache-2.0

Files in the repo

Repository payload23 top-level entries
  • .github
  • cli
  • lib
  • plugins
  • scripts
  • site
  • skills
  • tests
  • tools
  • web-ui
  • .coderabbit.yaml
  • .gitattributes
  • .gitignore
  • cli.js
  • LICENSE
  • package-lock.json
  • package.json
  • pnpm-lock.yaml
  • pnpm-workspace.yaml
  • README.md
  • README.vi.md
  • README.zh.md
  • web-ui.html

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More tools

JuliusBrussee/
caveman

🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman

105k
1 add
MemPalace/
mempalace

The best-benchmarked open-source AI memory system. And it's free.

59k
stablyai/
orca

Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop, mobile and remote runtime.

66k

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io

132k

Never stop coding. Free MIT AI gateway: one endpoint, 352 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/PWA. Built by 550+ contributors

64k
headroomlabs-ai/
headroom

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

71k