fx is Vercel's tiny, open-source coding agent built to get out of your way. Written in Zig and shipped as a ~6MB native binary, it starts almost instantly while keeping memory and context overhead low. Use it with local or cloud models, extend it with skills, plugins and MCPs, or embed it into your own agent infrastructure. Small by design, so the model gets more room to work.
Loading comments…
Project Info
Product Keywords
fx is Vercel's tiny, open-source coding agent built to get out of your way. Written in Zig and shipped as a ~6MB native binary, it starts almost instantly while keeping memory and context overhead low. It's a coding agent harness and CLI designed for research, automation, and embeddability as part of larger systems—not a heavy "IDE in the terminal" TUI, but something closer to a Unix shell in form and feel.
fx ships as a single native binary around 6.39MB, designed for instant installation and embedding in resource-constrained environments and agent sandboxes. No dependency hell, no bloated runtime—just a compiled tool that works.
Cold starts in 10µs with no unnecessary I/O before accepting user input. This makes fx exceptionally responsive for both interactive use and automated pipelines where every millisecond counts.
Optimal fx.wasm builds produced by the Zig toolchain further reduce size and make the network stack pluggable. You can run fx in browsers with WebAssembly JSPI support, like Safari 27+ and Chrome, with networking delegated to browser fetch.
Minimal system prompt and tools save on token costs and yield optimal time-to-first-token performance (TTFT). The design philosophy is simple: small by design, so the model gets more room to work.
Small by design, so the model gets more room to work.
fx's entire architecture revolves around this principle. Instead of piling on features, it strips everything to the essentials—minimal system prompt, minimal tools, minimal memory overhead—so the underlying model can focus on what matters. It's model- and provider-agnostic, working with local models, gateways, direct provider API access, or subscriptions. The shell-like UI preserves scroll history by default and produces minimal output, making it feel more like a Unix tool than a flashy TUI.
You're tired of heavyweight coding agents that eat RAM and add latency, or you're building agent infrastructure that needs a small, embeddable core. fx is experimental (v0.0.4, use at your own risk), but if you value minimalism, instant startup, and Unix-like extensibility, it's worth a look. It's also open source under Apache-2.0, so you can inspect every line and adapt it to your needs.
Other tools you might consider
Clears is an agentic execution platform for autonomous software delivery, helping R&D organizations move beyond individual AI tools to agentic execution across the entire SDLC.
GitHits gives coding agents access to the open-source code your app depends on. Get real implementation examples, dependency source navigation, package inspection and documentation. Agents can grep and read your codebase. They can't grep and read the open-source code your app depends on. That's where they start guessing, retrying, and looping. GitHits builds a version-aware index on demand. Agents can search, navigate, and inspect the code behind their dependencies. CLI: npx githits@latest init
You're running more coding agents than ever, but you can't keep up with them. That's where AgentPeek comes in. It pulls every session up into your Mac notch, live. Glance up, approve a prompt, watch token usage and manage the entire flow without pausing your YouTube video. All local, all yours.
Meet Mellum, a family of fast language models, including a next-generation model for ultra-low-latency and high-performance inference.
Maker
kettle_dev
Alternatives
Loading comments…