AgentToolDrop

7 of 53

opencodex

Local proxy that lets Codex and Claude Code run on Gemini, Grok, DeepSeek, Ollama or other LLMs

  • score 63
  • 17.5k stars
  • 146 stars/day since creation
  • pushed
  • v2.86.0 on
  • MIT

Install with npm

npm install -g @bitkyc08/opencodex   # Node 18+; the Bun runtime is bundled automatically
ocx start                         # proxy + dashboard on localhost:10100

Copied from the project's docs at github.com/lidge-jun/opencodex

Why you'd want it

opencodex is a local proxy that translates Codex's Responses API into whatever the provider speaks, covering streaming, tool calls, reasoning tokens and images. The README says it works with Codex, Claude Code, Claude Desktop and Grok Build, with 40+ built-in providers plus any OpenAI-compatible endpoint. A web dashboard on localhost:10100 is where you add providers, pick models and manage accounts, and it shows a live request log. Models are targeted with provider/model syntax such as codex -m "ollama/llama3", and combos give one virtual model id with failover or weighted round-robin across providers. ocx stop restores Codex to its original configuration.

Best for

  • Running Codex on non-OpenAI models
  • Using Claude Code with other providers
  • Failover across providers with combos

Use it when

You want Codex or Claude Code to use another provider's model, including a local Ollama model, and you are fine running a local proxy with a dashboard.

Skip it if

The README warns that some providers, notably Anthropic, may suspend or restrict accounts that route traffic through third-party proxies, and says the project is independent and not affiliated with OpenAI or Anthropic. It handles provider API keys, OAuth logins and optionally a pool of ChatGPT accounts, and ocx init wires Codex to the proxy, so read AGENTS_INSTALL.md for what it writes to disk. The desktop app is labelled beta and the Cursor adapter experimental; the from-source route pipes a curl installer for Bun into bash, which the npm route avoids.

Editor's note

Other routes in the README: desktop app (beta), Docker Compose, install from source. Docs at opencodex.me.

Related tools

Same category or shared tags.

  1. Ollama

    Model/API Standard

    Runs open models locally behind a CLI and a REST API, with an OpenAI-compatible endpoint

    • 183k stars
    • 152 stars/day since creation
    • pushed
    • v0.40.5 on
    • MIT

    Install on macOS or Linux

    curl -fsSL https://ollama.com/install.sh | sh
    Use it when, skip it if, stats for Ollama

    Why you'd want it

    Ollama downloads and runs open models such as Gemma with one command (ollama run gemma4) and serves them on a local REST API at localhost:11434. It supports a subset of the OpenAI API, so OpenAI clients work by changing the base URL, and it ships official Python and JavaScript libraries. ollama launch connects local models to coding agents including Claude Code, Codex, Copilot CLI and OpenCode, and the same API can reach Ollama's cloud models with an API key.

    Best for

    • Running open models on your own machine
    • Offline or private agent development
    • Pointing coding agents at local models

    Use it when

    You want a local model server that existing OpenAI-style clients and coding agents can talk to with minimal setup.

    Skip it if

    You only use hosted models from providers like Anthropic or OpenAI. Its OpenAI compatibility covers a subset of the OpenAI API, so check the docs if you depend on less common endpoints.

    Stats

    Stars per day, since repo creation
    152
    Release
    v0.40.5,
    License
    MIT
    Pricing
    Open source
    Works with
    macOS, Linux, Windows, Docker, Claude Code, Codex
    Forks
    18,323
    Open issues
    4,308
    GitHub checked

    Full stat block for Ollama

  2. Rank 12

    Omnigent

    Framework

    Orchestrates Claude Code, Codex, Cursor and other agents in one session, with policies and sandboxing

    • score 57
    • 10.8k stars
    • 85.4 stars/day since creation
    • pushed
    • v0.18.0 on
    • Apache-2.0

    Install with uv or pip

    uv tool install omnigent        # or: pip install "omnigent"
    Use it when, skip it if, stats for Omnigent

    Why you'd want it

    Omnigent is a meta-harness that puts a common orchestration layer over Claude Code, Codex, Cursor, OpenCode, Hermes, Pi and agents you write yourself, so you can swap or combine harnesses without rewriting. Agents are short YAML files with a prompt, local Python functions, MCP servers and sub-agents, and several agents can share one session, for example one reviewing another's work. Policies can pause for your approval before risky actions, cap spend or limit which tools an agent reaches, at server, agent or session level. Sessions can be followed from the terminal, a browser or a phone, and can run in cloud sandboxes such as Modal, Daytona or E2B.

    Best for

    • Running Claude Code and Codex in one session
    • Approval and spend policies across agents
    • Sharing live agent sessions with a team

    Use it when

    You want one layer to run and supervise several coding agents, with approval policies and sessions you can share or pick up from another device.

    Skip it if

    The README badges it alpha and says anonymized telemetry is collected by default, with an opt-out documented. It picks up model credentials from your environment, runs agents that execute commands on your machine, and lets a co-attached teammate's messages execute there too. On native Windows the Job Object backend does not isolate the filesystem or network.

    Stats

    New and rising score
    57 of 100
    Stars per day, since repo creation
    85.4
    Release
    v0.18.0,
    License
    Apache-2.0
    Pricing
    Open source
    Works with
    Claude Code, Codex, Cursor, macOS, Linux, Windows, Web, Docker
    Forks
    1,755
    Open issues
    2,143
    GitHub checked

    Full stat block for Omnigent

  3. Rank 22

    Munder Difflin

    Devtool

    Desktop app that runs Claude Code, Codex and other agent CLIs as a team coordinated by one agent

    • score 47
    • 8.8k stars
    • 63.8 stars/day since creation
    • pushed
    • v0.5.5 on
    • MIT

    Build from source

    git clone https://github.com/chaitanyagiri/munder-difflin.git
cd munder-difflin
npm install        # postinstall rebuilds node-pty against Electron's ABI
npm run dev        # launches the Electron app with hot reload
    Use it when, skip it if, stats for Munder Difflin

    Why you'd want it

    Munder Difflin is an Electron desktop app that wraps terminal agent CLIs such as Claude Code and Codex, each running as a real process in its own pseudo-terminal. You brief one orchestrator agent, which the README calls Michael or the GOD agent, and it assigns work and routes messages between agents through mailboxes in a shared hive of plain files. Spend, scope and destructive operations are escalated to an approvals queue for you, and the README says each agent has configurable autonomy and a circuit breaker for loops. Agents keep markdown memory with an optional semantic recall index, and per-agent git worktrees are available as optional isolation. It uses the subscriptions you already have for the CLIs, with bring-your-own keys and local models as options.

    Best for

    • Running several Claude Code and Codex agents in parallel
    • Briefing one orchestrator instead of many agents
    • Watching and steering agent terminals from one window

    Use it when

    You already use agent CLIs such as Claude Code or Codex and want several running on your own machine, coordinated by one orchestrator agent with human approval for spend and destructive operations.

    Skip it if

    The README calls it pre-release software, and it is an Electron desktop app that needs at least one agent CLI already installed, not a library or a headless tool. It runs real agent CLIs as processes you configure the autonomy of, stores provider keys and custom secrets, accepts inbound webhooks, and its dictation feature asks for Microphone and Accessibility permissions on macOS, so review the autonomy and permission settings before use. Official builds send anonymous usage events (opt out in Settings, with DO_NOT_TRACK, or by building from source), and the Pro tier with the Stapler and new sidebar is paid.

    Stats

    New and rising score
    47 of 100
    Stars per day, since repo creation
    63.8
    Release
    v0.5.5,
    License
    MIT
    Pricing
    Free tier + paid
    Works with
    Claude Code, Codex, Gemini CLI, Cursor, macOS, Linux, Windows
    Forks
    1,220
    Open issues
    254
    GitHub checked

    Full stat block for Munder Difflin

For agents

The same entry, as files an agent can read without scraping this page. 

# read one entry
curl -s https://agenttooldrop.com/tools/opencodex.md

# list the MCP servers in the catalog
curl -s https://agenttooldrop.com/tools.json \
  | jq -r '.tools[] | select(.category=="mcp-server") | .slug'