Vercel Open-Sources durable-harness for Crash-Proof Coding Agents
Vercel has released durable-harness, an open-source package that keeps coding agent sessions alive through crashes and timeouts.Vercel durable-harness wraps existing harnesses, including Pi, OpenCode, Claude Code and Codex, without changing them.Each session lives in a durable log.
A dead function no longer means lost work.For teams moving agents off laptops and into the cloud, that closes a real gap.TL;DR Size: npm package, version 0.0.x, Apache-2.0.Adapters cover 5 harnesses: Pi, OpenCode, fx, Claude Code, Codex.Runs on: Vercel Functions plus Vercel Queues, Node.
js 22+ with Postgres, or laptop files.Claude Code and Codex need a Vercel Sandbox.Performance: No latency or throughput benchmarks published.All recovery claims are self-reported by Vercel.Best: Finished model and tool calls are read back from the log, never re-run or billed twice.
Worst: Experimental 0.0.x API.Claude Code and Codex run only in a Vercel Sandbox, even self-hosted.What is durable-harness?durable-harness is a set of adapters from Vercel that store a harness session in a Workflow SDK World.A World bundles storage, a queue and a stream behind one interface.
The package uses the same machinery that powers Vercel’s Workflow SDK.Why do coding agents lose work in the cloud?Harnesses like Pi and Claude Code were designed as long-lived local processes.They keep the session in memory.
Locally, the workspace and history sit on the same disk, so resuming takes 1 prompt.In the cloud, a VM with persistent storage does not track which calls completed.Vercel team notes a failed turn or network error then means lost work.How does durable-harness work?
It writes each step of a turn to a durable log.Before a model or tool call runs, it records an intent.When the call finishes, it records the outcome.Each turn is backed by a message on Vercel Queues.If the function dies, the queue hands the turn to a new function.
That function reads the log and continues.After a crash, the log shows what finished, what was running and what never started: Completed calls are read from the log, so finished model tokens are not billed again.replay: 'safe' tools, such as lookups, can run again.
replay: 'never' tools, such as refunds, never run twice.The model is told the call may have happened.Harness session files are mirrored into the log as well.That covers Claude Code’s transcript, Codex’s rollout and OpenCode’s event store.Where do files and shell commands run?
Model-chosen commands never run on the function’s own disk.The tools docs list 3 workspace options: virtual() keeps files in the session log, with a bash-compatible shell and no network.It holds about 500 files.virtual({ durable: false }) keeps scratch files in memory and clears them every turn.
sandbox() gives each session a Vercel Sandbox, with /workspace on a Drive.Typed JSON documents, such as a to-do list, also live in the log.Each holds at most 64 KB.Run turn Kill function Reset
Press Run turn, then kill the function while CI is running. Scenario from Vercel's post: an agent adds a pricing page, opens a PR, and waits for CI. Illustrative simulation, not a benchmark. Click a tool, then crash it mid-call.
get_orderreplay: 'safe'
A lookup. Safe to repeat after a crash.
refund_orderreplay: 'never'
A refund. Must never run twice. Crash mid-call
Pick a tool and press Crash mid-call.
{ name: 'get_order', replay: 'safe', ... }
virtual()
Files in the session log. Bash-compatible shell, no network.
virtual({durable:false})
In-memory scratch. Cleared every turn.
sandbox()
Per-session Vercel Sandbox, /workspace on a Drive. Vercel
Vercel World + Agent Runs traces
Laptop
Local files, zero config
Self-host
Postgres World, Node.
js 22+ HarnessPackageRuns inStatus Pidurable-harness/pi-durableYour function + virtual workspace or SandboxSupported OpenCodedurable-harness/opencodeYour function + virtual workspace or SandboxSupported fxlibfxDurable on its ownSupported Claude Codedurable-harness/claude-codeVercel Sandbox onlyExperimental Codexdurable-harness/codexVercel Sandbox onlyExperimental Claude Code and Codex are single-user CLIs that keep state on disk, so each session gets its own Sandbox.
Source: vercel.com/blog/making-every-harness-durable · Vercel docs · Oct 2026© Marktechpost (function(){ var R=document.getElementById('mtp-dh'); function $(s
Related
相關文章
騰訊內部的“AI搜索配方”,公開了!
作者 | 畢偉豪 編輯|漠影 把一個AI時代很常見的問題輸進搜索框——查查騰訊最近的股價、產品和成員變動。這是谷歌的非AI搜索結果: ▲谷歌搜索結果 在谷歌的傳統搜索中,用戶得到的是一系列網頁鏈接;而在元寶搜索中,同樣的問題呈現出的結果卻大不相同。
谷歌聯手Meta奇襲英偉達,開源TorchTPU,給PyTorch代碼搬家
(公眾號:zhidxcom) 編譯 | 楊京麗 編輯 | 李水青 10月21日消息,今日,谷歌宣佈開源TorchTPU,這是谷歌和Meta PyTorch團隊共同開發的原生TPU軟件棧,開發者可以用較少代碼改動,將現有PyTorch模型部署到谷歌TPU上。
OpenAI 收購 AI 製藥初創 Orchestra Bio,把"管研發"的 Agent 收編進生命科學版圖
它收購了專注藥物研發項目 AI 協同管理的初創公司 Orchestra Bio,將團隊和技術一併整合進旗下的生命科學研究平臺 Rosalind。Orchestra Bio 的 CTO Andy Xu 擁有北大教育背景,這家公司的主打方向,是用 AI Agent 承接藥物研發全流程裡的任務梳理、進度跟蹤、風險預判、跨團隊協作這些項目管理工作——換句話說,它不替科學家做實驗,而是替團隊把"怎麼把藥研發出來"這件極其複雜的事管起來。
生數科技Motubrain團隊將直播Physical AI訓練全流程,持續半個月
預訓練:觀察訓練規模對模型性能及零樣本泛化能力的影響。2、後訓練:針對具體任務進一步訓練,提升任務執行能力。3、Agent任務執行:探索模型能否將已有經驗遷移到其他任務。4、RSI(遞歸自我改進)探索:研究模型能否根據執行反饋評估並改進自身行動。
IDC最新報告:阿里雲佔中國汽車公共雲市場份額超45%
本文作者: 徐咪 2026-10-21 17:28 導語:根據國際數據公司(IDC)最新發布的《中國汽車雲市場(2026年上半年)跟蹤》報告,2026年上半年中國汽車公共雲基礎設施市場規模約83.4億元人民幣。其中,阿 根據國際數據公司(IDC)最新發布的《中國汽車雲市場(2026年上半年)跟蹤》報告,2026年上半年中國汽車公共雲基礎設施市場規模約83.
AI日報:ChatGPT上線鎖定聊天功能;Ming-3.0-tts登頂國際評測雙榜首;Mistral 發佈 Voxtral 3 語音 Agent SDK
OpenAI 重磅上線 ChatGPT“鎖定聊天”功能:四位數 PIN 碼與生物識別雙保險OpenAI推出的“鎖定聊天”功能是隱私保護層面的實用改進,通過PIN碼與生物識別雙重驗證機制,有效防範設備外借場景下的隱私洩露風險。該功能設計較為剋制,鎖定對話及記憶信息完整保留,不影響原有使用體驗。