A self-hosted local coding agent that adds planning, candidate scoring, sandboxed testing and repair around compact open GGUF models.
co-l/openfox
A local-LLM-first agentic coding assistant that runs against vLLM, sglang, ollama or llamacpp backends with a browser UI.
Open-source alternative to:Claude Code
Cursor
- Stars
- 370
- Forks
- 78
- Open issues
- 131
- Last commit
- 3 days ago
What is it?
OpenFox is an autonomous coding agent built to run on local LLMs rather than hosted APIs. On first launch it detects a running local backend (vLLM, sglang, ollama or llamacpp) that exposes an OpenAI-compatible API and configures itself, then serves a web UI for working on the current project.
Work is organized around acceptance criteria. A builder mode implements tasks after an interactive planning step, and the agent keeps iterating until all criteria pass. Declarative workflows chain agent turns, sub-agents, shell commands and user approval gates, and an LSP integration gives feedback on code validity while edits are made.
The project also supports plugins that add LLM providers, tools, slash commands and skills, and it can use a vision model fallback so non-vision models can handle images. It requires Node.js 24 or later and is installed via npm.
Key features
- 01Detects local LLM backends such as vLLM, sglang, ollama and llamacpp automatically
- 02Plan-then-build workflow driven by acceptance criteria
- 03Declarative multi-step workflows with sub-agents and shell steps
- 04LSP integration for immediate code validity feedback
- 05Plugin system for adding providers, tools and slash commands
- 06Real-time metrics for prefill time, generation speed and context usage
Use cases
- Running a coding agent fully offline on your own GPU server
- Using an ollama or llamacpp setup for private code that cannot leave your machine
- Testing open-weight coding models against real project tasks
- Automating multi-step build and verification pipelines with user approval gates
Quick start
- 1
Install OpenFox globally with npm
npm i -g openfox - 2
Start OpenFox; it auto-detects your local LLM backend
openfox
Things to know
- You must already run a local LLM backend (vLLM, sglang, ollama or llamacpp) with an OpenAI-compatible API.
- Requires Node.js 24.0.0 or newer.
In collections
Topics
FAQ
- What is openfox?
- A local-LLM-first agentic coding assistant that runs against vLLM, sglang, ollama or llamacpp backends with a browser UI. OpenFox is an autonomous coding agent built to run on local LLMs rather than hosted APIs. On first launch it detects a running local backend (vLLM, sglang, ollama or llamacpp) that exposes an OpenAI-compatible API and configures itself, then serves a web UI for working on the current project.
- Is openfox free to use? What license does it have?
- openfox has no declared license. Its code is public, but without a license you have no explicit permission to reuse it; ask the author before relying on it.
- Is openfox still maintained?
- Actively maintained. The last commit to openfox was on Nov 4, 2026.
- What can I use openfox for?
- Running a coding agent fully offline on your own GPU server Using an ollama or llamacpp setup for private code that cannot leave your machine Testing open-weight coding models against real project tasks Automating multi-step build and verification pipelines with user approval gates
- How do I install openfox?
- Install OpenFox globally with npm: npm i -g openfox Start OpenFox; it auto-detects your local LLM backend: openfox
- Where can I get openfox?
- The source code and installation instructions are on GitHub: https://github.com/co-l/openfox
Summary generated by AI from the project's README. GitHub stats refresh daily.
Related projects
A desktop coding agent (Tauri app plus CLI) that runs on your existing AI subscriptions or local models, with built-in code search and verification checks.
A coding agent built on pi, tuned for small local LLMs, with background jobs, plan mode and per-phase model selection.
Open-source AI coding agent for VS Code that works with your own models, including hosted APIs, accounts, and local Ollama or llama.cpp.
An open-source coding agent for the terminal that runs on local models via Ollama or any OpenAI-compatible API.
A terminal coding agent written in Go that edits files, runs commands and keeps local sessions, with you choosing the model and permission level.