MCP & IDE Support

CloudAEye, right where you code.

Keep your editor. Bring deeper review, codebase context,
and better pull request decisions into your coding agent.

Connect your IDE Explore the commands
Cursor/CloudAEye
You

Review my pending changes for bugs and security issues before I open a pull request.

CloudAEye MCPreviewWorking-tree changes
Your coding agent calls CloudAEye.
The findings return to your conversation.
A second look, with the codebase in view.
High

A retry can charge the customer twice.

Concurrent requests can both read an unpaid status before either updates it.

payments/retry.ts:14
Medium

Untrusted input reaches the redirect.

Validate the return URL before passing it to the response.

auth/callback.ts:28
Choose which findings to address, then ask for a fix plan.
Review evidence, not automatic permission to merge.

Your IDE.
The right connection.

The review capabilities stay consistent. Installation, authentication, and skill selection follow your editor and coding agent.

Cursor

CloudAEye plugin + Agent chat

Cursor setup guide

  1. Install the plugin

    Add CloudAEye's plugin repository through Customize > Plugins, then reload Cursor.

  2. Authenticate CloudAEye

    Enable the plugin's MCP connection. Complete browser sign-in and consent; its OAuth bridge needs Node.js/npm.

  3. Choose a skill

    In Agent chat, use the / picker for init, then describe or review.

VS Code

CloudAEye plugin + GitHub Copilot Agent mode

VS Code setup guide

  1. Install from source

    Run Chat: Install Plugin From Source, add CloudAEye's repository, and approve the plugin's trust prompt.

  2. Connect the tools

    Use MCP: List Servers to start CloudAEye and authenticate. Enable its tools in Copilot Agent mode.

  3. Choose a skill

    Use the / picker to select init and describe. Copilot login and CloudAEye login are separate.

IntelliJ IDEA

JetBrains AI Assistant + MCP and imported skills

IntelliJ IDEA setup guide

  1. Prepare AI Assistant

    Enable JetBrains AI Assistant, import CloudAEye's skills, and install the runtimes required by the guide.

  2. Connect to CloudAEye

    Add the documented STDIO bridge in AI Assistant's MCP settings and complete CloudAEye authentication.

  3. Start an agent chat

    Activate Claude Agent or Codex. Pass custom MCP servers to the agent and select the installed skill with $ or Try in chat.

PyCharm

JetBrains AI Assistant + MCP and imported skills

PyCharm setup guide

  1. Prepare AI Assistant

    Enable JetBrains AI Assistant and import CloudAEye's skills. Node.js/npm is required for the OAuth bridge.

  2. Authenticate the connection

    Configure the bridge in AI Assistant's MCP settings, sign in to CloudAEye, and verify its tool list.

  3. Start with your change

    Activate Claude Agent or Codex, then choose init and describe from the installed skill picker.

CloudAEye hosted MCPhttps://api.cloudaeye.com/mcp

Use your IDE's guide. Registering MCP alone does not install skills; avoid adding a duplicate connection when the plugin supplies one.

Start with a Git repository, CloudAEye Code Review access, and an activated coding agent. Check the prerequisites.

From the first question
to the final review.

Understand, inspect, improve, and coordinate. Find the command for your task, with a clear scope and a concrete result.

Skills & Commands Coding Agent User Guide

These are CloudAEye skill names, not universal slash commands. Select them using your IDE's skill picker; names and invocation syntax depend on the installed agent and plugin.

Catch bugs before the PRinspect

Find logic errors, missed edge cases, concurrency problems, and invalid inputs in AI-generated or human-written changes. Inspect is the bug pass; use Security for vulnerabilities.

Scope
Pending changes, a path, or an open PR
Result
Findings with code locations and explanations
Inspect reference
Review the security surfacesecurity

Check application vulnerabilities, LLM and AI-agent integrations, MCP risks, and exposed secrets. Give security-sensitive changes a dedicated pass.

Scope
Pending changes, a path, or an open PR
Result
Security findings across five report types
Security reference
Take a comprehensive second lookreview

Combine the bug and security passes across all six report types before opening a significant pull request.

Scope
Pending changes, a path, or an open PR
Result
A combined review and its supporting evidence
Review reference
Check the work against the requirementscheck-task

Compare the implementation with a Jira ticket, GitHub issue, or written specification. Surface missing acceptance criteria before the change moves to review.

Scope
Pending changes and a task description
Result
Requirement coverage and implementation gaps
Check Task reference
Spot AI code that needs more scrutinyslop-score

Assess nine quality signals, including duplication, nonexistent APIs, dead code, missing tests, and overengineering. Use the evidence to focus manual review.

Scope
Pending changes, a path, or an open PR
Result
A quality score with contributing findings
Slop Score reference
Apply consistent PR quality standardscheck

Run the repository's checklist for documentation, tests, README freshness, dependencies, PR metadata, duplicate code, and potential secrets.

Scope
An open PR and its configured checklist
Result
A detailed checklist report posted on the PR
Check PR reference
Understand an issue before writing codeexplain

Connect a Jira or GitHub issue to the components it affects, their current behavior, and surrounding implementation. Start an unfamiliar task with code context.

Scope
A referenced issue or ticket
Result
A code-grounded explanation of the work
Explain reference
Ask a specific question about the changeask

Trace a question through callers, definitions, and usage in the repository's code graph. Ask for an explanation rather than a review verdict.

Scope
Pending changes and your question
Result
An answer grounded in repository context
Ask reference
Make your changes easy to understanddescribe

Summarize pending modifications and their impact across files for a PR description or commit message. Describe explains the change; it does not issue a review verdict.

Scope
Uncommitted changes
Result
A Markdown change summary
Describe reference
Follow the changed execution pathscontrol-flow

Visualize changes to function calls and execution paths. Identify affected callers and downstream behavior before merging a complex change.

Scope
A change and its code-graph context
Result
A sequence diagram of changed control flow
Control Flow reference
See the architectural impactreview-arch

Inspect how changes affect system architecture, repository layers, and dependencies. Bring cross-component risks into the review conversation.

Scope
Pending changes or an open PR
Result
Architecture diagrams and design findings
Architecture Review reference
Turn findings into a practical fix planimplement

Choose findings from a CloudAEye review and get a plan covering the files, affected callers, related defects, and tests. Your coding agent edits only when asked.

Scope
Selected findings from a previous review
Result
A structured plan, not an automatic code edit
Implement reference
Give new code the tests it needsadd-tests

Generate unit tests for uncovered code, including new test files when needed. Review and run the suggested tests before accepting them.

Scope
An open pull request
Result
Test suggestions on the PR, not local edits
Add Tests reference
Fill the documentation gapsadd-docs

Identify undocumented code and propose docstrings where they are missing. Accept only the suggestions that accurately explain your implementation.

Scope
An open pull request
Result
Documentation suggestions on the PR
Add Docs reference
Prioritize changes that need human reviewrisk-score

Assess PR risk using bug history, rework, code ownership, author history, and AI slop score. Understand why a change deserves closer attention.

Scope
An open PR and its engineering history
Result
A risk report to guide review priority
Risk Score reference
Find the right people to reviewassign-reviewer

Consider ownership, contribution history, review queues, availability, and urgency. Read the recommendations, then explicitly request reviews from the people you choose.

Scope
An open pull request
Result
Reviewer recommendations and relevant context
Assign Reviewer reference
Put business-critical PRs firsttriage-prs

Prioritize a queue by business impact, complexity, and risk. Bring your organization's OKRs into the decision rather than simply reviewing by arrival time.

Scope
A pull request queue and business context
Result
Priorities with a rationale for review order
Triage PRs reference
Give review follow-ups an ownercreate-issue

Turn selected findings into Jira tickets or GitHub issues with an assignee and context. Identify findings already tracked in the session to reduce duplicate work.

Scope
Selected review findings
Result
Real tracking issues in your chosen system
Create Issue reference
Establish who knows the codecreate-owners

Use Git contribution history to establish an ownership starting point in an unfamiliar repository. Refine the result with your team's current responsibilities.

Scope
A repository or chosen folder
Result
OWNERS.md files written into the repository
Create Owners reference
Prepare your repository for reviewinit

Detect the provider, remote URL, and base branch. Configure the monitor branch and complete provider integration when requested.

Scope
The current Git repository
Result
Repository configuration for CloudAEye workflows
Init reference
Find the command and options you needhelp

Explore the available command catalog, arguments, options, examples, and limitations. Check your installed version before starting an unfamiliar workflow.

Scope
The catalog or a named command
Result
Usage guidance without running the command
Help reference

A path-scoped review covers that path, not the entire change. PR commands have eligibility requirements; consult the reference for scope, side effects, and available options.

Put the commands
to work together.

One-off checks are useful. Connected workflows take you from understanding the task to making an informed review decision.

Before the first line of code.

Explain the issue, identify relevant components and owners, then check your implementation against the original requirements.

explaincreate-ownerscheck-task

After the review finds a gap.

Plan the fix, add test and documentation suggestions, and track the findings that need follow-up. Review and run suggestions before accepting them.

implementadd-testsadd-docscreate-issue

A large diff needs a clear story, too.

PR walkthrough organizes changes into logical groups, explains their dependencies, and highlights risks and architectural impact.

Explore PR walkthrough

More capable agents.
Developers still in control.

CloudAEye supplies context, evidence, and suggestions. Your team chooses what to change, whose review to request, and whether the code is ready to merge.

Read the workflow guide
  • Accept suggestions deliberately. Inspect proposed docs and tests, and run tests before accepting them.
  • Know which actions write. Some commands post PR reports, create tickets, or write ownership files.
  • Review with the right scope. A clean result is evidence for the reviewed change, not a blanket guarantee.

Before you connect.

Is MCP the same as installing the plugin?

No. MCP connects your coding agent to CloudAEye's tools. Skills guide the workflow, including preparing the change for review. Cursor and VS Code use the packaged plugin; JetBrains integrations use MCP and imported skills. Follow your IDE's guide for both parts.

Do all IDEs use the same command syntax?

No. Cursor and VS Code expose installed skills through the / picker. The JetBrains guides use the installed skill picker or Try in chat. Use the names shown by your client instead of assuming a universal prefix. Codex and Claude Code have their own Codex and Claude Code pages.

Does my code stay entirely on my machine?

No. Pre-commit describes when the review happens, not where it runs. The workflow uploads the change to CloudAEye for analysis. Review your organization's policies and the documented data flow before connecting a private repository.

Can I review before creating a pull request?

Yes. Inspect, Security, and Review support pending changes, with optional path scope. Add Tests, Add Docs, and other PR workflows require an open pull request. The User Guide explains eligibility and limitations.

What if authentication or a review fails?

Agent login and CloudAEye authentication are separate. Check your IDE guide, the hosted endpoint, enabled tools, and account access. A failed or expired analysis is not a clean review; investigate the error rather than treating it as approval.

Stay in your editor.
Bring the full review workflow with you.

Choose your IDE Browse Skills & Commands

Using a coding agent directly? Explore Codex or Claude Code.