AI coding tools have accelerated the code-generation process, but the rest of the engineering cycle — reviewing changes, monitoring systems, and handling routine maintenance — hasn’t kept the same pace. And that gap is where the latest release from Cursor is aimed.
Cursor is the AI-native code editor developed by startup Anysphere, and last week it introduced Automations, a feature that lets teams run coding agents automatically in response to events or schedules. The goal is to push AI beyond the editor and into the day-to-day mechanics of maintaining a codebase.
Cursor casts agents into the background
Automations allows developers to define triggers and instructions that launch a cloud-based coding agent whenever certain events occur. Those triggers can include signals from common engineering tools such as GitHub, Slack, or Linear, as well as custom webhooks defined by the user.
Inside Cursor, developers can create these automations directly from the agent dashboard, selecting a repository, branch, and task for the agent to carry out. Once configured, the system can launch a cloud agent that runs against the codebase and performs the requested action, such as auditing security issues or investigating bugs first reported in Slack.
Or else, a team can configure an automation to monitor PagerDuty alerts and automatically investigate incoming incidents, triaging the issue against the codebase and opening a pull request with a proposed fix where appropriate.
Once triggered, Cursor spins up a cloud sandbox where the agent executes the task using whatever models and tools the team has configured. From there it can analyze the repository, run tests, generate patches, or comment on pull requests before handing the results back to developers.
The idea is to automate parts of the engineering process that typically occur after code has been written. Teams might configure agents to review pull requests for security issues, triage bug reports overnight, or compile regular summaries of activity across a repository.
Cursor says agents can also retain information between runs through a memory tool, allowing them to learn from previous investigations and handle recurring issues more efficiently.
“Automations can store and retrieve memories, enabling them to accumulate context across runs and become more effective over time,” the company wrote in a blog post.
The company argues that the surge in AI-generated code has increased development output across engineering teams, while review and maintenance processes have struggled to keep pace. Automations aims to close that gap by allowing agents to monitor development systems continuously and respond to events as they occur, rather than waiting for a developer to initiate the task manually.
Smooth background operators
The release reflects a broader shift underway in AI-driven development tools. Early coding assistants focused on generating snippets or completing lines of code inside an editor. The current wave of tools is pushing agents into longer-running tasks across the engineering stack.
GitHub, for instance, is experimenting with agentic workflows that can investigate CI failures, update documentation, or produce repository reports automatically inside GitHub Actions. At the same time, a growing set of tools is emerging to coordinate these longer-running systems, with developers increasingly supervising multiple agents that work across a codebase in parallel. New model releases are also targeting longer-running development tasks. Anthropic’s Claude Opus 4.6 and OpenAI’s GPT-5.3 Codex, for instance, are designed to handle extended coding sessions and multi-step engineering work rather than short prompts.
Cursor’s Automations feature fits squarely within this shift. By allowing agents to run continuously in response to events across development tools, the company is positioning its agents as background operators reacting to activity in development systems. Instead of being invoked directly in an editor session, an automation may run whenever a new issue is created or a pull request is merged.
That structure essentially treats coding agents as participants in the development process itself. A team might configure an agent to watch for certain patterns in pull requests, verify fixes for recurring bugs, or monitor production alerts and prepare a patch before a developer begins investigating.
The model still assumes human review before changes reach production, but the agent carries out much of the investigative and drafting work that would normally precede a fix.
Multi-repo support is coming
Some early discussion around the launch centred on how the system determines which development environment an automation should run in. One user asked: “I might be missing it. How does it know which environment to use when you have multiple repos?”
Members of the Cursor team responded that developers configure the repository and runtime environment when setting up an automation. “You get to pick the repo/environment to use. And you can set up your environment to have really anything installed in the VM!” one team member replied.
This exchange highlights one of the current limitations of the feature. At launch, automations are scoped to a single repository, meaning teams running large multi-repo codebases would need to configure separate automations for each project.
However, this limitation may be temporary. Another member of the Cursor team clarified this in a follow-up response. “Automations are scoped to a repo right now (so you can just have your automation running separately on each repo). Multi-repo support in one automation coming soon!,” they said.
Cursor pushes beyond its own editor
Automations arrives alongside another notable piece of expansion news. Cursor also announced that its agents can now run inside JetBrains IDEs through the Agent Client Protocol (ACP), which enables tools like IntelliJ IDEA, PyCharm, and WebStorm to access Cursor agents.
The move signals that the company is trying to reach developers who remain embedded in established development environments rather than migrating entirely to Cursor’s own editor.
Together, the announcements suggest a strategy that extends Cursor’s agents across more of the software development lifecycle. Automations introduces always-running agents that monitor and act on engineering signals, while the JetBrains integration widens the number of environments where those agents can be used.
For teams experimenting with AI-native development, the underlying idea is straightforward: if coding agents are producing more of the code, they may also need to help manage the growing stream of pull requests, alerts, and maintenance work that follows.
