- Published on
Cursor Projects β Coding Agents Now Watch Slack and File PRs on Their Own
The old limitation of coding agents was simple: they only worked when you were sitting at your desk. Cursor is changing that.
On October 13, 2026, Cursor rolled out Projects in beta to all users. The centerpiece is a coordinator agent that doesn't wait for a prompt. It watches Slack channels, runs on a defined schedule, or reacts automatically whenever a pull request is opened.
Projects: The Agent Moves First
Traditional AI coding tools followed a question-and-answer pattern: you ask, the agent responds. Projects flips this dynamic.
Once the coordinator agent is configured:
- Slack monitoring: It reads bug reports or feature requests in a Slack channel and modifies the relevant code
- Scheduled execution: It runs daily stand-ups, weekly reviews, or pre-deploy checklists on your timetable
- PR follow-through: It reads open pull requests and automatically handles downstream tasks
From an edtech operations standpoint, this isn't a convenience feature β it's a structural shift. Imagine an agent that catches a content error a teacher posted in Slack at night and has a fix PR ready before the school day starts.
Cloud Agent Canvases: Share Results as a Link
On October 16, another capability landed: cloud agents can now produce interactive canvases β and share them with teammates via a single link.
Code review summaries, test reports, data analysis β the agent's work finally has a shareable format. The same update raised the repository limit to 300 and improved loading speed.
An Environments API also launched in beta, letting developers programmatically create and manage the environments their agents run in.
iOS Remote Control: Direct Agents From Your Phone
The October 6 update brought Remote Control to the iOS app. Without opening a laptop, developers can view and direct agents running on their own machine from an iPhone. It works without cloud agents and is enabled by default for most users except Enterprise organizations.
Reviewing an agent mid-task from a cafΓ©, redirecting it, and letting it finish β that's now possible.
The Next Stage of Vibe Coding
If vibe coding was about generating code with natural language, Projects shifts the paradigm to agents that choose their own moment to act. Here are practical scenarios:
| Situation | Projects in Action |
|---|---|
| Customer feedback drops in Slack | Agent finds the relevant code and suggests a fix |
| Every Monday morning | Agent auto-generates a summary of last week's PRs and tech-debt report |
| A PR is opened | Agent runs lint and tests, then posts review comments automatically |
One caution: because the coordinator agent can open PRs on its own judgment, teams should define clearly what scope it operates in and what approval steps remain human. The convenience of automation shines brightest when the guardrails are firmly in place.
Sources: