DigestAI news desk

Cut through the AI noise.

Agents & Tools3 min read

Vercel open-sources durable-harness to keep coding agent sessions alive after crashes

Vercel released durable-harness, an open-source npm package (version 0.0.x) under Apache-2.0 that stores each step of a coding‑agent turn in a durable log, allowing sessions to survive crashes and timeouts. The package wraps existing harnesses—Pi, OpenCode, fx, Claude Code, Codex—without changing them.

1 source

Key points

  • Vercel released durable‑harness, an open‑source npm package (version 0.0.x) that logs coding‑agent steps to survive crashes.
  • Adapters cover five harnesses—Pi, OpenCode, fx, Claude Code, Codex—and run on Vercel Functions, Queues, Node.js 22+ with Postgres or locally.
  • Finished model and tool calls are read from the log, preventing duplicate token billing; benchmarks are not published and the API is experimental.

The log records intent before a model or tool call and outcome after it finishes. If a function dies, Vercel Queues hand the turn to a new function, which reads the log and continues. Finished calls are read back, so tokens are not billed twice, and safe tools can be replayed while irreversible actions are not repeated. The package runs on Vercel Functions plus Queues, on Node.js 22+ with Postgres, or locally; Claude Code and Codex require a Vercel Sandbox. No latency or throughput benchmarks have been published; performance claims are self‑reported and the API is experimental.

By closing the gap where cloud‑based coding agents previously lost work after crashes, durable‑harness aims to make moving agents from laptops to the cloud more reliable and cost‑effective.

Full story from MarkTechPost · by Michal SutterOpen source ↗

Vercel Open-Sources durable-harness for Crash-Proof Coding Agents

MarkTechPost · 21 October 2026

Loading the full article…

This text was published by MarkTechPost and written by Michal Sutter. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗

Topics · follow one to build your own front page

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.

Comments

via GitHub Discussions

More in Agents & Tools

All →

Related stories