Cursor previously offered its Automations feature—essentially agents that run background tasks—on a separate web page at cursor.com/automations. With the 3.5 release on May 20, that page has been integrated directly into the Agents Window within the IDE, placing the code-writing interface and the task-running interface side by side.
Three Changes, One Key Upgrade
The latest version of Automations introduces three changes:
- Moved from a web dashboard to a built-in IDE panel
- Supports attaching multiple repositories to a single automation
- Allows automations to run without any repository, acting as a tool listener
The first change is about product entry—engineers will use it more frequently. The most significant is the second: cross-repository support.
Anyone working with microservices knows that a single change often touches three repos: frontend, backend, and infrastructure. Previously, the agent could only handle one repo at a time, requiring cross-repo tasks to be broken into three steps, with the agent losing context between steps. This update allows an automation to attach multiple repos simultaneously, letting the agent review all relevant code before acting.
Cursor's official changelog highlights use cases: cross-repo PR reviews, test fixes, and build runs. This directly addresses the biggest pain point for mid-to-large engineering teams.
Five Repository-Free Templates: What's the Play?
The third change may seem marginal but is worth examining.
"Repository-free automations"—agents that run in Cursor without reading any of your code—signal that Cursor is positioning itself as a hub for engineers' daily tools, not just a code-writing IDE.
Five official templates make this clear:
- Slack Digest: Automatically summarizes missed DMs and channels each morning.
- Product Analytics: Weekly report template that pulls data from data warehouses like Databricks.
- Product FAQ: Monitors Slack Q&A channels and answers using documentation and codebase.
- Product Finance: Pulls invoices and refunds from billing tools like Stripe.
- Customer Health: Monitors anomalies across Granola, Slack, and Databricks.
A code-writing IDE is now taking over tasks for PMs (data analysis), sales (Stripe), and customer success (Slack monitoring).
The most interesting template is the Product FAQ Agent, which monitors Slack Q&A channels and answers questions by referencing documentation and code. For small teams, this is like hiring a 24/7 junior engineer who never complains about repetitive questions.
Competitive Landscape
Looking at the past three weeks:
- Early May: Anthropic released Managed Agents on Code with Claude (multi-agent orchestration, Dreaming, and Outcomes).
- Mid-May: Google Antigravity 2.0 launched a similar agent cluster at $100/month.
- Cursor's move bets on the idea that agents can not only write code but manage entire workflows.
All three are competing to transform agents from "tools that run commands in an IDE" into "operational schedulers for entire engineering teams." Cursor's advantage is that it's already the window most engineers have open daily—no need to switch apps, giving it a natural moat.
Cursor also launched a 7-day promotion: 50% off agent runs for new automations. The short promotion period suggests the goal is not to attract new users with discounts but to lock existing users into workflows before Anthropic and Google can.
What to Watch Next
In the short term, two metrics matter:
- The number of user-built automations on the Cursor Marketplace—if a batch of repository-free templates emerges within a month, it indicates real adoption.
- The average completion time for cross-repo PRs—a hard metric for whether microservice teams are actually using the feature.
As for how far IDE vendors can go in taking over entire workflows: Slack started as a messaging tool and gradually absorbed calendars, documents, and automation. The difference is that Cursor has one card Slack didn't: agents.
Sources: Cursor 3.5 changelog (cursor.com); CocoLoop; Cursor Releases Automations Platform for AI Coding Agent Management (mlq.ai)