Top 200 ReposEXPLORER
Browse agents, categories and pages
Back to results
Reported by sourceThe source says this happened. Top 200 Repos has not reproduced it.

Hand coding tasks to Cursor cloud agents that open pull requests

Start Cursor agents from the editor, web, Slack or a GitHub comment; each works in its own cloud VM on a branch and returns a pull request with screenshots, videos and logs.

The problem

How it works

Cursor’s documentation describes cloud agents (formerly background agents) that clone a connected GitHub, GitLab, Bitbucket or Azure DevOps repository into an isolated VM, work on a separate branch and push the result for review. You can start them from Cursor Desktop, cursor.com/agents, the iOS app, Slack, Linear, a @cursor comment on an issue or pull request, or the API, and run many in parallel. The docs put most of the weight on the environment: a snapshot, agent-led setup or a Dockerfile in .cursor/environment.json so the agent can install dependencies, run tests and use MCP servers.

Results and limitations

Cursor says agents return merge-ready pull requests with screenshots, videos and logs of how they checked their work, and that you can take over the agent’s remote desktop to try the change without checking out the branch.

Before you try it

Cloud agents need a Premium, Teams or Enterprise plan (not Free or Plus), bill at the chosen model’s rate, and ask for a spend limit on first use. Unlike the foreground agent they auto-run terminal commands with internet access by default, which Cursor itself flags as a prompt-injection and data-exfiltration risk; restrict outbound domains and use runtime secrets, which are redacted from transcripts and commits. Results depend on your setup and were not independently reproduced by Top 200 Repos.

Tools and repositories

No ranked repository is involved. This use case still stands on its own source.

Tools
Cursor · GitHub · Docker
Prerequisites
A Cursor plan that includes cloud agents and a source-control connection set up by an account admin.

Try it yourself: Brief a cloud agent on one scoped task

Fill the fields to adapt the prompt. Empty fields stay as visible [placeholders].

Preview
Work on [Repository] on a new branch named [Branch name]. Task: [Task]. Before changing code, read the relevant files and list your plan in the pull request description. Success means [Success criteria]; run the test command [Test command] and include its output. Do not touch files outside [Allowed paths], do not change CI configuration or secrets, and do not fetch any URL that is not a package registry or official documentation. Open a draft pull request with screenshots or logs that show how you checked the change, and stop there for my review.

Set up the environment and network restrictions first; an agent that cannot run the tests cannot verify its own work.

Sources

  1. Cloud Agents (opens in a new tab)

    Project documentation · cursor.com · Undated · Source checked 11 Nov 2026, 10:02 UTC

    What this source supports
    • Agents run in isolated cloud VMs, work on a branch and push changes
    • Start from desktop, web, iOS, Slack, Linear, GitHub comments or the API
    • Environment setup via snapshot, agent-led setup or .cursor/environment.json
    • Pull requests come with screenshots, videos and logs; remote desktop takeover
    • Requires Premium, Teams, Enterprise or a legacy paid plan; spend limit on first use
  2. Cloud Agents: Secrets and Network (opens in a new tab)

    Project documentation · cursor.com · Undated · Source checked 11 Nov 2026, 10:02 UTC

    What this source supports
    • Cloud agents auto-run terminal commands and have internet access by default
    • Cursor names prompt injection and data exfiltration as risks
    • Runtime secrets are redacted from transcripts, commits and tool results
    • Network egress can be restricted per team or environment
Report a problem

Planned, not done

Run coding agents from one self-hosted control centre with OpenHands

OpenHands Agent Canvas runs the OpenHands agent, or Claude Code, Codex, Gemini CLI and other ACP agents, on your machine or in Docker sandboxes, and can trigger them from schedules or webhooks.

github.comOpenHands/OpenHands
Reported by source

Build an app with Replit Agent while keeping production data out of reach

Replit Agent builds and browser-tests apps from a description. A widely reported 2025 incident, where it deleted a live database during a code freeze, shows why separate dev and prod data matters.

docs.replit.com
Reported by source

Explore and change a large codebase from the terminal with Gemini CLI

Gemini CLI is Google’s open-source terminal agent. Simon Willison found it strong on large codebases in 2025; in 2026 Google moved individual users to Antigravity CLI.

github.comgoogle-gemini/gemini-cli
Reported by source

Rebuild a side project in OpenCode to test provider lock-in

Matthew Brunelle rebuilt a YouTube Music to OpenSubsonic shim with OpenCode and Kimi K2.6 to compare it with an earlier Claude Code and Opus 4.6 attempt.

blog.matthewbrunelle.com · 30 Apr 2026anomalyco/opencode
Reported by source

Ship a non-trivial app feature over many short agent sessions

Mitchell Hashimoto built Ghostty's unobtrusive macOS update notifications with the Amp agent over 16 sessions, rewriting and reviewing parts by hand before merging.

mitchellh.com · 11 Oct 2025