Get an independent review of a task from another Claude session
A user sends a Claude chat URL into a Cowork task so the agent can review and check the work done in that chat. ·
Done withClaude CoworkClaude
- What they did
- The user opens the same task in Claude chat in a browser and in Claude Desktop’s CoWork feature, then gives CoWork the URL of the browser chat. CoWork, with permission to use the browser and local tools, is told to bounce questions, outputs, and recommendations off the other Claude session and ask that agent questions before coming back to the person.
- How it went
- They say the two agents “ping pong” through the task so many issues get ironed out before handoff, which significantly reduces manual review. In their follow-up, they estimate it cuts about 90% of back-and-forth and say it works on almost all tasks.
- Worth knowing
- The poster says there is no fancy setup required beyond Claude Desktop’s CoWork tab and a Claude chat URL.
Try it yourself with Claude Cowork
Here is the URL of a Claude chat where I did some work: [chat URL]. Review the work in that chat independently: check the facts, calculations, and reasoning, and flag anything wrong, missing, or unsupported. Give me a short list of issues with suggested fixes, and don't change any files or send anything until I approve.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
Source: Reddit · Jul 1, 2026
Related use cases
Build a C compiler with a team of parallel agents
Anthropic's engineering write-up on tasking Opus 4.6 agent teams with building a C compiler.
Recreate a video ad as a JavaScript animation
The author had Claude recreate a downloaded ad video as a pure JavaScript animation with the original audio, working autonomously with ffmpeg in under an hour.
Dispute a car rental damage charge with Claude
A traveler used Claude as an agent to contest a Sixt damage charge; it advised accusing the company of timestamp fraud and complaining to a regulator, which the article says backfired.
Run a solo company with eight agent departments
A founder set up CEO, CFO, COO, lawyer, accountant, marketing, CTO, and improver agents that share memory and consult each other.
Run a commercial real estate business with an AI agent
Drew McAllister's agent flagged counteroffers by text, built owner follow-up lists, tracked bids, drafted emails, read lease redlines and flagged deadline conflicts.
Custom agent
Run browser agents for job applications, logins, and data fetching
A poster tested six browser agents on daily tasks like job applications, logins, data fetching and posting; Manus was versatile but tripped bot detection and cost too much.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.