• Tech Dev NotesTech Dev Notes
Apps
  • App lookup
  • App compare
Market movement
  • App charts
  • App rankings
Visual proof
  • App screens
  • App listing screenshots
  • App icons
Build intelligence
  • App tech stacks
  • Tool releases
  • Developers
More
  • X feature flags
  • Grokipedia
  • Blog
  • Follow on X
Skip to content
All content/ filesChangelog

cursor-docs/latest/content · Jun 26, 20:20 UTC

pages/cloud-agent/capabilities.txt

TXT·4.3 KB·42 lines

content/

  • pages

    • api.txt
    • approval-agents.txt
    • bugbot.txt
    • cloud-agent.txt
    • customize-cursor.txt
    • enterprise.txt
    • evals.txt
    • extension-api.txt
    • hooks.txt
    • mcp.txt
    • models-and-pricing.txt
    • plugins.txt
    • rules.txt

  • pages/account

    • regions.txt
    • update-access.txt

  • pages/account/enterprise

    • billing-groups.txt
    • cyber-safeguards.txt
    • service-accounts.txt

  • pages/account/organizations

    • organization-admin-api.txt

  • pages/account/pricing

    • request-based-legacy.txt

  • pages/account/teams

    • admin-api.txt
    • ai-code-tracking-api.txt
    • analytics-api.txt
    • analytics.txt
    • dashboard.txt
    • members.txt
    • pricing.txt
    • scim.txt
    • setup.txt
    • sso.txt

  • pages/agent

    • agent-review.txt
    • agents-window.txt
    • debug-mode.txt
    • design-mode.txt
    • overview.txt
    • plan-mode.txt
    • prompting.txt
    • security.txt

  • pages/agent/security

    • run-modes.txt

  • pages/agent/tools

    • browser.txt
    • canvas.txt
    • search.txt
    • terminal.txt

  • pages/cli

    • acp.txt
    • changelog.txt
    • github-actions.txt
    • headless.txt
    • installation.txt
    • mcp.txt
    • overview.txt
    • shell-mode.txt
    • using.txt

  • pages/cli/reference

    • authentication.txt
    • configuration.txt
    • output-format.txt
    • parameters.txt
    • permissions.txt
    • slash-commands.txt
    • terminal-setup.txt

  • pages/cloud-agent

    • automations.txt
    • best-practices.txt
    • capabilities.txt
    • choose-runtime.txt
    • my-machines.txt
    • private-workers.txt
    • security-network.txt
    • self-hosted-cloud-run.txt
    • self-hosted-k8s.txt
    • self-hosted-pool.txt
    • settings.txt
    • setup.txt

  • pages/cloud-agent/api

    • endpoints.txt
    • webhooks.txt

  • pages/configuration

    • worktrees.txt

  • pages/configuration/migrations

    • vscode.txt

  • pages/cursor-review

    • merge-queue.txt
    • overview.txt
    • pr-inbox.txt
    • pr-page.txt

  • pages/cursor-review/cli

    • authentication.txt
    • parameters.txt
    • stacking.txt

  • pages/customizing

    • aws-bedrock.txt

  • pages/enterprise

    • baa.txt
    • compliance-and-monitoring.txt
    • deployment-patterns.txt
    • endpoint-security.txt
    • identity-and-access-management.txt
    • llm-safety-and-controls.txt
    • model-and-integration-management.txt
    • network-configuration.txt
    • organization-groups.txt
    • organizations.txt
    • pooled-usage.txt
    • privacy-and-data-governance.txt
    • private-connectivity.txt
    • security-hardening.txt

  • pages/get-started

    • quickstart.txt

  • pages/integrations

    • azure-devops.txt
    • bitbucket.txt
    • cursor-blame.txt
    • github.txt
    • gitlab.txt
    • jetbrains.txt
    • jira.txt
    • linear.txt
    • microsoft-teams.txt
    • notion.txt
    • slack.txt
    • xcode.txt

  • pages/mcp

    • install-links.txt

  • pages/models

    • claude-4-5-haiku.txt
    • claude-4-5-sonnet.txt
    • claude-4-6-sonnet.txt
    • claude-4-sonnet-1m.txt
    • claude-4-sonnet.txt
    • claude-fable-5.txt
    • claude-opus-4-5.txt
    • claude-opus-4-6-fast.txt
    • claude-opus-4-6.txt
    • claude-opus-4-7-fast.txt
    • claude-opus-4-7.txt
    • claude-opus-4-8.txt
    • cursor-composer-1-5.txt
    • cursor-composer-1.txt
    • cursor-composer-2-5.txt
    • cursor-composer-2.txt
    • gemini-2-5-flash.txt
    • gemini-3-1-pro.txt
    • gemini-3-5-flash.txt
    • gemini-3-flash.txt
    • gemini-3-pro-image-preview.txt
    • gemini-3-pro.txt
    • glm-5-2.txt
    • gpt-5-1-codex-max.txt
    • gpt-5-1-codex-mini.txt
    • gpt-5-1-codex.txt
    • gpt-5-2-codex.txt
    • gpt-5-2.txt
    • gpt-5-3-codex.txt
    • gpt-5-4-mini.txt
    • gpt-5-4-nano.txt
    • gpt-5-4.txt
    • gpt-5-5.txt
    • gpt-5-codex.txt
    • gpt-5-fast.txt
    • gpt-5-mini.txt
    • gpt-5.txt
    • grok-4-20.txt
    • grok-4-3.txt
    • grok-build-0-1.txt
    • kimi-k2-5.txt

  • pages/reference

    • deeplinks.txt
    • ignore-file.txt
    • keyboard-shortcuts.txt
    • permissions.txt
    • plugins.txt
    • sandbox.txt
    • third-party-hooks.txt

  • pages/sdk

    • python.txt
route: /docs/cloud-agent/capabilities
title: Capabilities
description: What Cloud Agents can do.

Capabilities
Computer use
Each cloud agent runs in its own isolated VM with a full desktop environment. Agents can use a mouse and keyboard to control the desktop and browser, allowing them to interact with the software they build like a human developer.
This means agents can start dev servers, open the app in a browser, click through UI flows, and verify their changes work before pushing a PR. Read more in the announcement blog post.
Demos and Artifacts
Agents create artifacts such as screenshots, videos, and log references to demonstrate their work. These artifacts are attached to the PR so you can quickly validate changes without checking out the branch locally.
Artifacts in GitHub
You can opt-in to have Cloud Agents embed artifacts directly into GitHub pull request descriptions by enabling the Allow posting artifacts to GitHub setting in the Cloud Agents dashboard.
GitHub's image proxy requires public URLs, so artifacts in PR descriptions use long, unguessable URLs that are viewable without authentication. For context, GitHub used public URLs for all issue and PR attachments until May 2023.
Remote desktop control
You can take control of the agent's remote desktop to interact with the software the agent is building. Hand control back to the agent at any time to let it keep working.
Cloud agents run in a remote VM that can be fully onboarded with your repo, dependencies, tooling, and setup scripts. This allows you to test changes directly in the agent's VM without checking out the branch on your local machine.
MCP tools
Cloud agents can use MCP (Model Context Protocol) servers configured for your team. This gives agents access to external tools and data sources like databases, APIs, and third-party services during their runs.
Add and enable MCP servers through the MCP dropdown in cursor.com/agents.
Cloud agents support OAuth for MCP servers that need it. OAuth is per-user, including for MCP servers shared at the team level.
Custom MCP servers
You can add custom MCP servers using either HTTP or stdio transport. SSE and mcp-remote are not supported.
MCP configurations are encrypted at rest. Sensitive fields are redacted and cannot be read back by any user after saving:
env — environment variables for stdio servers
headers — request headers for HTTP servers
CLIENT_SECRET — OAuth client secret for HTTP servers
HTTP vs stdio
HTTP (recommended) — server configurations are never present in the cloud agent's VM environment. The agent does not have access to refresh tokens, headers, or other credentials. Tool calls are proxied through the backend.
Stdio — servers run inside the cloud agent's VM, so the agent has access to the server's configuration and environment variables. This is similar to how stdio MCPs work in the Cursor IDE.
Stdio servers depend on the VM environment to execute. We cannot verify that a stdio server will run successfully until a cloud agent is launched. We recommend using HTTP MCPs when possible, and configuring your environment setup correctly if you use stdio servers.
Fixing CI Failures
Cloud Agents automatically try to fix CI failures in PRs they create. This currently supports GitHub Actions only.
Cloud Agents skip automatic CI follow-ups if:
You've pushed a new commit to the branch; cloud agents do not auto-fix CI failures on human commits.
You've sent a follow-up message to the agent.
The same check is already failing on the base commit of the PR.
The PR has already had 10 CI-failure follow-ups.
To disable this feature on all your personal Cloud Agents, go to Cursor Dashboard → Cloud Agents → My Settings and disable the "Automatically fix CI Failures" option.
To disable this feature on a specific Cloud Agent PR, you can comment @cursor autofix off on the PR. To re-enable it, comment @cursor autofix on.
If you want cloud agents to fix CI failures in your own PRs, you can simply ask them by tagging Cursor in a comment as normal. For example, @cursor please fix the CI failures, or @cursor fix the CI lint check failure.
Automatically fixing CI failures is currently only available on Teams; support for non-Teams accounts is coming soon. In the meantime, if you want similar behavior, you can ask the cloud agent explicitly to monitor and fix CI failures on the PR.
Previouspages/cloud-agent/best-practices.txtNextpages/cloud-agent/choose-runtime.txt

© 2026 Tech Dev Notes

RSSAboutAPIPrivacyTermsSitemap@techdevnotes