Skip to content

GitHub PR Review Agent

Deploy an AI agent that reviews pull requests and posts comments — callable via a single API endpoint.

What You’ll Build

An agent endpoint at /api/run/pr-reviewer that:

Accepts a PR URL or diff as input
Uses GitHub MCP tools to read files, check CI status
Analyzes code quality, security, and best practices
Returns structured review comments

Step 1: Create the Agent Endpoint

Dashboard → Agent Endpoints → Create

{
  "name": "PR Reviewer",
  "slug": "pr-reviewer",
  "system_prompt": "You are an expert code reviewer. Analyze the provided code changes and provide actionable feedback. Focus on: bugs, security issues, performance, and readability. Be specific — reference line numbers and suggest fixes.",
  "workspace_id": "your-workspace-id",
  "llm_provider": "anthropic",
  "llm_model": "claude-sonnet-4-6",
  "output_format": "json",
  "output_schema": {
    "type": "object",
    "properties": {
      "summary": { "type": "string" },
      "issues": {
        "type": "array",
        "items": {
          "type": "object",
          "properties": {
            "severity": { "type": "string", "enum": ["critical", "high", "medium", "low"] },
            "file": { "type": "string" },
            "line": { "type": "number" },
            "message": { "type": "string" },
            "suggestion": { "type": "string" }
          }
        }
      },
      "approved": { "type": "boolean" }
    }
  },
  "enable_streaming": true
}

Step 2: Add GitHub MCP to Your Workspace

Add the GitHub MCP server to your workspace so the agent can access PR data:

Browse MCP Marketplace → GitHub
Install to your workspace
Add your GitHub token as a workspace secret

Step 3: Call the Agent

curl -X POST https://api.aerostack.dev/api/run/pr-reviewer \
  -H "Authorization: Bearer aek_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"input": "Review PR #42 in repo myorg/myapp"}'

Response (JSON output format):

{
  "output": {
    "summary": "3 issues found. The auth middleware change has a potential bypass.",
    "issues": [
      {
        "severity": "critical",
        "file": "src/middleware/auth.ts",
        "line": 45,
        "message": "JWT verification skipped when header is empty string",
        "suggestion": "Change `if (!token)` to `if (!token || token.trim() === '')`"
      }
    ],
    "approved": false
  },
  "usage": {
    "tokens_input": 2400,
    "tokens_output": 800,
    "cost_cents": 12,
    "latency_ms": 3200
  }
}

With SSE Streaming:

const response = await fetch('https://api.aerostack.dev/api/run/pr-reviewer', {
  method: 'POST',
  headers: {
    'Authorization': 'Bearer aek_your_api_key',
    'Content-Type': 'application/json'
  },
  body: JSON.stringify({ input: 'Review PR #42 in repo myorg/myapp' })
});

const reader = response.body.getReader();
const decoder = new TextDecoder();

while (true) {
  const { done, value } = await reader.read();
  if (done) break;

  const text = decoder.decode(value);
  // Parse SSE events: thinking, tool_call, tool_result, done
  for (const line of text.split('\n')) {
    if (line.startsWith('data: ')) {
      const event = JSON.parse(line.slice(6));
      console.log(event.type, event);
    }
  }
}

Step 4: Integrate with CI

Add to your GitHub Actions workflow:

- name: AI PR Review
  run: |
    REVIEW=$(curl -s -X POST https://api.aerostack.dev/api/run/pr-reviewer \
      -H "Authorization: Bearer ${{ secrets.AEROSTACK_AGENT_KEY }}" \
      -H "Content-Type: application/json" \
      -d "{\"input\": \"Review PR #${{ github.event.pull_request.number }} in ${{ github.repository }}\"}")

    echo "$REVIEW" | jq '.output.issues[] | "\(.severity): \(.file):\(.line) — \(.message)"'

Next Steps