一个想法,做成真的
设置昵称

← 作品广场

Fieldwork Code已上线

Run parallel AI coding sessions on your laptop or in the cloud, and pick them up from your phone
开发者工具工具:AI coding agents作者:Fieldwork (maker, Product Hunt)2026-11-18

Product overview

Fieldwork Code is an AI coding platform, inspired by Conductor and born out of building fieldwork.ai (per its Product Hunt page). It runs parallel coding sessions with your choice of harness — Claude Code, Codex, OpenCode, Pi, or Fieldwork's own harness — and your choice of model.

Target and users

Developers who run several agent sessions at once and want to start work at a desk, then check, steer and review sessions from the web or a phone while away from the machine.

Page structure

A chat conversation for the session, with a dedicated side pane holding a terminal, file diffs, PR status and CI check results. Sessions started locally or in the cloud stay available from the desktop app, the web, and the Fieldwork mobile app.

Core loop

Start a session locally (free) or in a cloud sandbox, let the chosen harness work, then review diffs in the side pane, annotate them, and send one-click actions back to the chat — push a PR, review the code, resolve conflicts — with customisable prompts.

Features and pricing

  • Local coding via the desktop app is free
  • Cloud subscription $50/mo gives 1,440 vCPU-hours and 2,880 GB-hours a month, equivalent to a 2 vCPU / 4 GB sandbox around the clock; larger boxes are also available (all figures per the Product Hunt page)
  • Cloud clients also get hosted apps, a datalake for connector-extracted data, and agent and automation capabilities

Feature details

Inline diff annotations can be sent to the coding agent as manual code reviews. Because it is part of the broader Fieldwork offering, it ships with 3,000+ connectors, a Linear/Jira-style task tracker and a Confluence/Notion-style knowledge base, both integrated with AI chat (per the launch page).

Visual and technical

Desktop app for local runs, cloud sandboxes for remote runs, plus web and mobile access to the same sessions. Harness- and model-agnostic by design: you pick the harness and the model per session.

Not included / boundaries

It is a session platform, not its own coding model — the actual coding is done by the harness you select. Local sessions depend on your own machine; always-on cloud capacity is the paid tier described above.

1VR Curated · Source: Product Hunt @Fieldwork

评论 (1)

并行开几个会话后用手机接着盯 diff 和 CI,太真实了。想问本地和云端切换时,上下文会丢吗?
全栈阿斌2026-11-22

在 X 分享 复制链接

+ 发布