fx

A terminal coding agent that works with local models through Ollama or cloud providers and supports embedding in editors and browser apps.

Alternative to:

fx is a coding agent for developers who prefer working in a terminal and researchers who want to embed an agent in their own software. Its shell-style interface keeps tool activity alongside the conversation and preserves terminal scrollback.

Model connections: Use Ollama or another OpenAI-compatible endpoint, OpenRouter, Vercel AI Gateway, or models accessed through ChatGPT and Grok subscriptions.
Interactive and scripted work: Work through an ongoing terminal conversation or send individual requests from scripts and CI, with a full transcript available for review.
Agent extensions: Add skills, MCP servers, custom tools and subagents; skill instructions and MCP tool definitions load when needed to limit context use.
Editor and app embedding: Connect to compatible editors through Agent Client Protocol, or embed the agent core and terminal through libfx, including WebAssembly applications.
Image input: Include images in requests or let the agent read image files alongside other project material.

The native CLI works with your own files and doesn't require a background service. Inference runs at your chosen endpoint: a local server processes requests locally, while a cloud provider receives the model requests. The browser demo uses a browser workspace; embedded applications control storage, tools and network access.

fx is written in Zig and is open source under Apache 2.0.