Skip to main content
Use Greptile’s MCP server to automatically retrieve and resolve code review comments directly in your IDE.
Prerequisites: First set up the MCP server in your IDE, then follow the workflows below.

​
Overview

Once you have set up the Greptile MCP server in your IDE, you will be able to:
  • Retrieve unaddressed review comments
  • Search for specific types of issues
  • Get context about why issues were flagged
  • Apply fixes based on Greptile’s suggestions
…all from within your IDE.

​
Retrieving Comments

​
Get Comments for Current PR

1

Identify Current PR

When working on a branch, your IDE will automatically detect the associated PR.
2

Ask for Comments

Use these proven prompts with Claude:
  • “Show me all unaddressed Greptile comments for this PR”
  • “What security issues has Greptile found recently?”
  • “Search for Greptile comments about error handling”
3

Review Comment Details

Claude will show you:
  • Complete comment text with context
  • File paths and specific line numbers
  • Whether comments are addressed or still open
  • Related custom context patterns that were violated
  • PR information and repository details

​
Search for Specific Issues

Find Comments by Type:
  • “Find all security-related Greptile comments in this repository”
  • “Show me performance issues flagged by Greptile”
  • “Get error handling comments from recent reviews”
Find Comments by File:
  • “Show Greptile comments for the current file”
  • “Are there any unaddressed reviews for src/auth.js?”
  • “What issues has Greptile found in this component?”
These searches work across your entire organization’s repositories and return real comment data with file locations and context.

​
Resolving Comments Step-by-Step

​
Automated Resolution

1

Select Comment to Fix

In your IDE, click on a Greptile comment or use: “Fix this Greptile comment: [paste comment text]”
2

Analyze the Issue

The IDE will:
  • Read the comment and understand the problem
  • Examine the code context
  • Check related custom context patterns
  • Determine the appropriate fix strategy
3

Generate Fix

The coding agent will:
  • Propose a specific code change
  • Explain why this fix addresses the issue
  • Show before/after code comparison
4

Apply Fix

Review and apply the suggested changes:
  • “Apply this fix”
  • “Show me the diff first”
  • “Explain why this fixes the issue”
5

Verify Resolution

Confirm the fix addresses the original concern:
  • Run relevant tests
  • Check for syntax errors
  • Validate the solution makes sense

​
Manual Resolution with AI Guidance

Get Detailed Explanation:
  • “Explain this Greptile comment in detail and suggest how to fix it”
  • “Why is this flagged as a security issue?”
  • “What’s the best way to resolve this performance concern?”
Get Fix Suggestions:
  • “How should I fix this memory leak issue?”
  • “Suggest 2-3 ways to address this code duplication”
  • “What’s the proper error handling pattern here?”

​
Practical Examples

​
Security Issue Resolution

Greptile Comment:
“Consider adding input validation to prevent potential security vulnerabilities. The user input should be sanitized before processing.”
In Claude Code: 
You: "Fix this security issue: Missing input validation for user ID parameter"

Claude Code: I'll help you add proper input validation. Looking at your code, I can see the getUserData function is missing validation. Here's the fix:

[Shows before/after code with input validation added]

Would you like me to apply this fix and also check for similar issues in other functions?

​
Performance Issue Resolution

Greptile Comment:
“N+1 query problem detected - consider using joins or batch loading”
In Cursor: 
You: "How do I fix this N+1 query problem?"

Cursor: I can see the issue in your code. You're making individual database queries in a loop. Here's how to fix it with a join:

[Shows optimized query with proper joins]

This will reduce your database calls from N+1 to just 1. Should I apply this change?

​
Code Style Violation

Greptile Comment:
“Use camelCase for variable names according to team guidelines”
Resolution Workflow: 
You: "Fix the naming convention issues in this file"

IDE: I found 3 variables that need to be converted to camelCase:
- user_data → userData
- is_valid → isValid
- api_response → apiResponse

I'll update these and any references. Applying changes now...

​
Advanced Workflows

The following sections cover advanced automation. For basic comment resolution, the steps above are sufficient.

​
Batch Comment Resolution

Resolve Multiple Comments: 
You: "Show me all unaddressed Greptile comments for this PR and fix the ones you can automatically"

IDE: I found 8 unaddressed comments:
✅ Fixed: 3 naming convention issues
✅ Fixed: 2 missing error handling cases
⚠️  Needs review: 1 architecture concern
⚠️  Needs review: 2 complex logic issues

I've applied the automatic fixes. The remaining 3 comments need your input because they involve design decisions.

​
Context-Aware Fixes

Using Custom Context: 
You: "Fix this error handling according to our team patterns"

IDE: Based on your organization's custom context, I see you use Result<T, E> for error handling. I'll update this function to follow that pattern:

[Shows conversion from try/catch to Result pattern]

This matches the pattern used in 15 other functions in your codebase.

​
Integration with Development Workflow

Pre-commit Resolution: 
You: "Before I commit, check if there are any unaddressed Greptile comments in my changes"

IDE: I found 2 unaddressed comments in your staged files:
1. Missing null check in user.service.js:45
2. Inconsistent error message format in auth.controller.js:23

Would you like me to fix these before you commit?

​
Best Practices

​
When to Use Auto-Fix

Safe for Auto-Fix:
  • Code style and formatting issues
  • Simple null checks and validation
  • Naming convention corrections
  • Basic error handling additions
Requires Review:
  • Architecture and design changes
  • Complex logic modifications
  • Performance optimizations with tradeoffs
  • Security fixes affecting business logic

​
Verification Steps

  • Code Review: Always review auto-generated fixes before applying them to ensure they make sense in context
  • Test execution: Run relevant tests after applying fixes to catch any regressions
  • Review integration: Mark comments as addressed only after verification
  • Rollback capability: Provide undo functionality for problematic fixes

​
Human Oversight

  • Confidence scoring: Only auto-apply high-confidence fixes
  • Review queuing: Queue uncertain fixes for human review
  • Explanation logging: Document what changes were made and why
  • Approval workflows: Require human approval for critical file changes
⌘I