Install ProPR, or skip the server
Run npm install -g propr-cli and propr setup on a Linux Docker host (quickstart). ProPR Hosted runs a dedicated instance for you when you would rather not run one.
ProPR: the self-hosted alternative to Cursor Cloud Agents
Cursor Cloud Agents run in Cursor's cloud VMs, started from the editor, web, Slack, Linear, the API, or a GitHub comment. ProPR centers on planned changes, GitHub pull requests, and plain-comment follow-up on your server.
vs
Cursor
Where the agent lives
Cursor starts from its own agent, models, and cloud. ProPR starts from the process around a planned change, with the agents you already subscribe to.
Cloud Agent pathEditor, web, Slack, or
@cursor → Cursor cloud VM → pull requestBugbot reviews the PR and Autofix spawns the fix.ProPR pathPlan → isolated workspace → GitHub PR → follow-up commitReview in GitHub and follow up with a plain comment, on your own server.
Switch from Cursor Cloud Agents to ProPR.
What you keep: your GitHub repositories and review process, Cursor itself for live editing, and any Claude or ChatGPT plan, which ProPR runs through Claude Code or Codex.
The same ideas, ProPR's names
| Cursor Cloud Agents | ProPR |
|---|---|
| Cloud Agent | Task |
| Plan Mode | Plan, approved before code runs |
@cursor on a PR or issue | Any plain PR comment, or the AI label on an issue |
| Bugbot and Autofix | /review and /fix; Ultrafix to a score |
| Cursor's cloud VM | Isolated workspace on your server |
| Cursor's model catalog, usage-billed | Your provider plans or API keys, no markup |
From zero to the first pull request
Add the same repository and sign in your agent
Open the web UI, sign in with GitHub, and add the repository. Sign in Claude Code, Codex, or Antigravity CLI with the plan you already hold.
Start the first task and approve the plan
Add the AI label to the issue you would have sent to a Cloud Agent, then approve the plan.
Review the pull request in GitHub
Review the diff the way you review any pull request. A plain comment on it becomes the next commit on the same branch; /review posts scored findings and /fix applies them. The follow-up playbook walks one comment to its commit.
What stays with Cursor Cloud Agents: ProPR has no editor, iOS app, or Slack entry point; tasks start from GitHub, the web UI, the desktop app, the CLI, or MCP.
Cloud Agents, renamed from Background Agents in October 2025, start from the editor, cursor.com/agents, the iOS app, Slack, Linear, the API, or an @cursor comment on a GitHub PR or issue, and cover a lot of the loop inside Cursor's product. They produce merge-ready pull requests, tag-in from a PR comment to fix CI failures, and subscribe to PR comments and reviews. Bugbot reviews every PR update, and Bugbot Autofix spawns a Cloud Agent to fix what it finds. The editor's Plan Mode writes an implementation plan before any code. Cursor supplies its own models alongside Anthropic, OpenAI, Google, and others, with third-party models billed by usage from the Premium tier at $30 per month and up. Since March 2026 a self-hosted worker can run the agent VMs in your infrastructure while orchestration and model access stay with Cursor.
ProPR puts the pull request at the center, on infrastructure you run end to end. The change starts as a reviewed plan, the agent works in isolation, the result becomes a pull request, and plain GitHub comments continue the change with no mention syntax. One process holds when some people use Cursor, others use a terminal agent, and everyone reviews in GitHub.
Many teams use both: Cursor for live editing and exploration, ProPR for planned, reviewable delivery.
Cursor Cloud Agents fit best when...
- Most AI coding already starts inside Cursor, and the same agent should continue in the cloud.
- Cursor's agent, model catalog, and Bugbot review cover the loop.
- Cursor hosting the orchestration and model access is acceptable.
ProPR fits best when...
- Plain GitHub comments drive follow-up commits from any of five agents.
- Planning, tasks, logs, and usage live on your own server, outside any single IDE.
- Claude Code, Codex, Antigravity CLI, Mistral Vibe, and OpenCode share one path on subscriptions you already hold.
Cursor-hosted agents versus a self-hosted PR path.
| Question | Cursor Cloud Agents | ProPR |
|---|---|---|
| Starts a task from | Editor, web, iOS, Slack, Linear, GitHub, API | Web UI, desktop app, GitHub issues, MCP, CLI |
| Plan before implementation | Plan Mode in the editor | Reviewed plan on every task |
| Opens the pull request | Yes | Yes |
| GitHub comments become follow-up commits | @cursor on the PR | Any plain comment, any agent |
| PR review | Bugbot, with Autofix | /review scores, /fix applies |
| Agents it runs | Cursor's agent on Cursor, Anthropic, OpenAI, Google models | Claude Code, Codex, Antigravity CLI, Mistral Vibe, OpenCode |
| Runs on your infrastructure | Self-hosted worker; orchestration and models at Cursor | Yes, Apache 2.0 |
| Platform pricing beyond model costs | Cloud agents from Premium, $30/mo; Teams from $30 per user | No |
Verified November 2026: cursor.com/docs/cloud-agent, Cloud Agent capabilities, Bugbot, Plan Mode, Cursor 2.0 changelog, self-hosted cloud agents, cursor.com/pricing.