libfx adds durable, resumable agent turns
- Source
- x.com
- Date
Harness durability in https://t.co/OL0LzGtvAw. The same 𝚌𝚛𝚎𝚊𝚝𝚎𝙵𝚡𝙰𝚐𝚎𝚗𝚝() is now *durable* on your laptop or in the cloud, via files, Postgres, or Vercel Workflows. Node.js or Bun. Agents need long-running compute. But they also need to tolerate restarts, redeploys, timeouts, crashes… Concretely, if a user sends a prompt and then throws their phone into a black hole, we ensure that turn runs to completion. If the process crashes mid-tool-call, it resumes from there. To this end, 𝚕𝚒𝚋𝚏𝚡 introduces a simple durability adapter API. On Vercel, it leverages all the infra we built for durable workflows and queues, giving you reliability with 1LOC. This is a critical aspect of production-grade agents (and something all @eve agents also get out of the box). Next, we'll show you how we bring the same durability to Pi, OpenCode, and even your own custom harnesses.
Context
fx is an open-source (Apache-2.0), model-agnostic coding- and command-line tool written in Zig, shipped as a single small binary and labeled experimental; libfx is its core packaged as a library, so another application can embed the and decide its own storage, tools and network access (fx.sh). Guillermo Rauch announced that agents created through libfx can now be made durable, on a laptop or in the cloud, under Node.js or Bun (X).
Durable here means an agent's turn, the stretch of work it does after receiving a prompt, including any tool calls, outlives the process running it. Rauch's examples: if a user sends a prompt and then disconnects, the turn still runs to completion, and if the process crashes in the middle of a tool call, it resumes from that point instead of starting over or being lost. The same holds across restarts, redeploys and timeouts. libfx does this through what Rauch calls a durability adapter API, where the developer picks the backing store: local files, Postgres, or Vercel Workflows. On Vercel, the post says, the adapter reuses the platform's existing durable-workflow and queue infrastructure and takes one line of code.
The evidence here is the announcement alone. The fx home page reviewed here does not mention the durability feature (its separate docs were not part of the material), and the post does not explain what state is saved at each step or how a half-finished tool call is handled on resume, so those details remain unconfirmed. Rauch says the same durability will next be brought to other harnesses, naming Pi and OpenCode, as well as custom ones.
- agent harness — The scaffolding around a model that turns it into a working agent — the loop, the tools it can call, and the rules for when to stop.
- AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Checking sign-in…
Loading comments…


