Skip to content

Summary

Vercel Labs' tiny open-source coding agent: a ~6MB Zig binary that cold-starts in microseconds and embeds anywhere.

Screenshots

Description

fx is a coding agent built the opposite way to the current crop of terminal IDEs. Written in Zig by Vercel Labs and released under Apache-2.0, it is a single binary of around 6-8 MB that cold-starts in roughly 10 microseconds and holds a single-digit-megabyte memory baseline — small enough to embed inside another system rather than run as an application you sit in front of.

Design

The interface is a Unix shell, not a full-screen TUI: output scrolls, history stays where you left it, and the agent composes with the rest of your shell instead of taking the terminal over. The stated audience is researchers and developers who want a harness they can drive and modify.

Capabilities
  • Model-agnostic — works against local models, direct API providers, the Vercel AI Gateway, an OpenAI ChatGPT subscription or xAI's Grok.
  • Permission modes for controlling what the agent may do without asking.
  • Skills, plugins and MCP servers as extension points, wired together on Unix composition principles.
  • Subagent delegation for splitting work.
  • Builds to WebAssembly as well as a native binary, so the same agent can run in a sandboxed host.
Install

curl -fsSL https://fx.sh/setup.sh | bash

Status

fx is explicitly experimental — the project is at an early version and Vercel Labs warns to use it at your own risk. It is free and open source (Apache-2.0), with about 2.2k stars on GitHub.

Who it is for

People who want to study or extend an agent loop rather than consume one, and teams looking for an agent binary light enough to ship inside a product, a CI job or a sandbox.

Reviews

Similar App Suggestions

App: Roomote

Roo Code

New

Self-hostable cloud coding agent from the Roo Code team — investigates repos, verifies its own work and opens pull requests, driven from Slack, Teams, Discord or Telegram rather than an IDE.

Coding & DevelopmentFreemium

An Apache-2.0 TypeScript framework for building AI agents — workflows, memory, RAG and evals — with a local studio and an agentic software factory on top.

Coding & DevelopmentFreemium

AI design engineer that generates distinctive UI designs and production code inside your own repo, Figma and design system.

Coding & DevelopmentFreemium

App: oMLX

Jun Kim

New

Menu-bar LLM inference server for Apple Silicon, with continuous batching and tiered KV caching that keeps local models fast enough for real coding work.

Coding & DevelopmentFree

Open-source local inference server that profiles your hardware, picks models that fit, and points your coding agent at them — free, private and offline.

Coding & DevelopmentFree

Covered in the Weekly