Cursor version of OpenClaw debuts! AI reviews code and fixes bugs on its own, programmers' lobster freedom is here?

AI Lobster for developers?

Early this morning.CursorAnnouncing the launch of OpenClaw-like features -CCursor Automations Features, allowing AI to automatically help developers review, monitor, and fix code 24/7, as well as assist in the R&D process.

Based on this feature, developers can simply configure the intelligences to continuously monitor and optimize the codebase to build theA truly automated “software development factory”.”. And, all of these automation capabilities are provided byCloud Agent Driver, they will use their own arithmetic to build, test and demonstrate the results of their work.

Atlasis, creator of the AI learning tool RRecallAI, says that at the rate this trend is going, “in the future we'll just have to be super awesome botmasters.” Another developer said he's been running a full-time online Agent through OpenClaw for a few months now, and that Cursor was smart to add this feature to the IDE. Still other developers exclaimed that this is crazy, Cursor is turning into OpenClaw 🦞. Because OpenClaw Logo and name are related to lobsters, so it is used to build Agent Assistant is also jokingly referred to as “lobster”.

These Agents can run on developer timed tasks or be triggered by various events such as sending a Slack message, creating a new Linear work order, merging GitHub PRs, having a PagerDuty alert, and more. Developers can also configure custom events via Webhook.

Currently Cursor has released a batch of automation Agents on its website, including 12 that fix CI failures, summarize changes daily, find vulnerabilities, generate documentation, and more.

Regarding pricing, Lee Robinson, head of education at Cursor AI, revealed in response to developer questions that theseThe cloud Agent uses the same token as in the editor, can enjoy the developer's Ultra subscription benefits.

However, some developers have questioned the security of automated Agents: who is to blame when an Agent merges a fix patch at 3am but messes with hanging production environments? As well as the lack of human involvement when autonomous intelligences make changes based on triggering conditions.

I. Improve code review, monitoring, and maintenance efficiency with automation capabilities

With the rise of programming intelligences, every developer is producing more code, but the efficiency of code review, monitoring and maintenance has not yet increased in tandem.

Cursor says its update formally leverages automation capabilities to improve the efficiency of these parts of the development lifecycle at scale.

When invoked, the Automation Agent starts the cloud sandbox, executes commands based on the MCP protocol and model configured by the developer, and self-checks the output.

The Agent can also use memory tools to learn from past executions and become more accurate the more it is used.

Over the past few weeks, the researchers have been running more automation Agents in Cursor's own codebase, resulting in two main types of automation scenarios:Code review and monitoring (Review and monitoring), daily affairs processing (Chores).

Second, all-day automatic review, monitoring, repair code, Cursor internal has deployed three major Agent

In terms of review and monitoring, automation can be applied to code change reviews, which can identify and fix code style and format inconsistencies, security vulnerabilities, performance degradation and other types of problems.

Cursor's automated code review Agent Bugbot runs when a PR is created or updated, is triggered thousands of times a day, and has found millions of vulnerabilities since it went live.

Its new automation features let developers customize various types of review intelligences for different scenarios. Here are the three currently used within Cursor:

Security reviewAutomation Agents are triggered every time code is pushed to the master branch. Intelligentsia can spend longer digging into more hidden and detailed issues without blocking the PR process.

It audits code differences for security vulnerabilities, automatically skips issues already discussed in the PR, and sends high-risk findings to Slack.Cursor says this automated feature has already captured multiple security vulnerabilities with serious flaws within Cursor.

Agentic codeowners (intelligent codeowners)Agent performs risk grading based on the scope of impact, complexity, and infrastructure impact at each PR creation or code push.

Low-risk PRs are automatically reviewed and approved; high-risk PRs are assigned up to two reviewers based on contribution history.

And the results of all its decisions are aggregated to Slack and logged to the Notion database via MCP so that developers can audit the intelligence's behavior and optimize commands.

Incident responseThe Automation Agent is triggered on PagerDuty alerts and launches an Agent to view logs, check for recent changes to the code base via Datadog MCP.

It then sends a message to the engineer on duty in the Slack channel, attaching monitoring information with a PR containing a fix, which can shorten the developer's failure response time.

Third, check the bugs themselves, report to the user early in the morning every day, push important changes every week

The researchers found that automation is also useful for routine tasks and knowledge work that requires stitching together information from different tools.

Weekly summary of changes's automated Agent, will push a weekly summary in Slack that summarizes important changes to the codebase over the past 7 days and highlights the major PRs, bug fixes, technical debt, security and dependency updates that have been merged.

Test coverageThe automated Agent will, each morning, review the recently merged code and identify the parts that need additional test coverage. It will write tests that follow the project's existing specifications, modifying the production code only as necessary.The Agent runs the relevant tests before creating and submitting a PR.

When vulnerability reports appear on the Slack channel.Bug report triageThe Automation Agent will first check if it is a duplicate issue and create a work order via Linear MCP. The Agent then troubleshoots the root cause in the code base and tries to fix it, and finally replies in the original thread with a summary of the processing.

Abhishek Singh, developer of Rippling, a US-based enterprise management platform, has built a personal assistant based on Cursor's automation features.

Singh would post meeting minutes, to-do lists, task lists, and links to Loom videos to a Slack channel throughout the day, have a timed task Agent run every two hours to read them, correlate his GitHub PRs, Jira tasks, and Slack mentions, de-duplicate the information from multiple sources, and then post a neat summarized Kanban board.

He also uses automation triggered by Slack to create Jira tasks from conversation threads and sync discussion summaries to Confluence. they have extended the automation to scenarios such as troubleshooting, weekly status reports, and shift handoffs.

Conclusion: AI Programming Moves Into Full-Link Automation, Needs to Be Wary of Security Risks

Today, OpenAI introduced GPT-5.4 for Codex to enable native computer usage, and Cursor launched an automation system for always-on programming, which has extended the application scenarios of AI programming to the field of R&D full-link automation. In this way, automation of the greatly improved programming process has led to a significant increase in the efficiency of small teams, and developers may play more of a role in the future in theAI AgentThe role of the manager.

But as many developers have said, the more authority an AI Agent has, the more blurred the boundaries of risk and responsibility become. When an intelligent body can autonomously write code, merge PRs, run tests, and even deploy on-line, once it misinterprets the requirements and buries a logic loophole.造The question of who should bear the losses and responsibilities has become one that the entire industry must face up to.

Source: Smart Stuff