Multi-model AI code review server using OpenRouter - get diverse perspectives from multiple LLMs in parallel
npm install @klitchevo/code-councilOne AI can miss things. A council of AIs catches more.
Code Council runs your code through multiple AI models simultaneously, then shows you where they agree, where they disagree, and what only one model caught.
``Consensus Analysis
Critical: SQL Injection Vulnerability
Location: src/api/users.ts:42
The user input is directly interpolated into the SQL query without sanitization.
Use parameterized queries instead.
---
High: Missing Input Validation
Location: src/api/users.ts:38
The userId parameter is used without validation. Add type checking.
---
Session Token Expiration
Location: src/api/auth.ts:28
- Kimi K2.5: "Tokens should expire after 24 hours"
- DeepSeek V3.2: "Current 7-day expiration is reasonable for this use case"
- Minimax M2.1: "No issue found"
---
Low: Magic Number
Location: src/utils/pagination.ts:12
Found by: GLM 4.7
The value 20 should be extracted to a named constant.
`
Different AI models have different strengths:
- One model might miss a security issue another catches
- Unanimous findings are almost certainly real problems
- Disagreements highlight where you should look closer
- Single-model findings might be noise, or might be the one model that saw something others missed
Think of it as getting 4 senior engineers to review your code at once.
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
`json`
{
"mcpServers": {
"code-council": {
"command": "npx",
"args": ["-y", "@klitchevo/code-council"],
"env": {
"OPENROUTER_API_KEY": "your-api-key-here"
}
}
}
}
Get your API key at OpenRouter.
That's it. Ask Claude: "Use review_code to check this function: [paste code]"
Run reviews directly from command line:
`bashReview git changes
npx @klitchevo/code-council review git --review-type diff
> More setup options: See Configuration Guide for Cursor, VS Code, custom models, and advanced options.
GitHub Actions
Automatically review PRs with multiple AI models. Findings appear as inline comments on the exact lines of code. Code fixes use GitHub's suggestion syntax for one-click apply. Re-runs automatically clean up old comments.
$3
Generate the workflow file automatically:
bash
npx @klitchevo/code-council setup workflow
`This creates
with inline PR comments enabled.Options:
-
--simple - Use markdown format instead of inline comments
- --force - Overwrite existing workflow file$3
Or create the workflow manually:
`yaml
name: Code Council Reviewon:
pull_request:
types: [opened, synchronize, ready_for_review, reopened]
jobs:
review:
runs-on: ubuntu-latest
if: github.event.pull_request.draft == false
permissions:
contents: read
pull-requests: write
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- uses: actions/setup-node@v4
with:
node-version: '20'
- name: Run Code Council Review
env:
OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
run: |
npx @klitchevo/code-council review git \
--review-type diff \
--format pr-comments \
> review.json
- name: Post Review
env:
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
run: |
gh api repos/${{ github.repository }}/pulls/${{ github.event.pull_request.number }}/reviews \
--method POST \
--input review.json
`Add
to your repository secrets (Settings > Secrets > Actions).Use Cases
| Scenario | Tool | What You Get |
|----------|------|--------------|
| About to merge a PR |
review_git_changes | Multi-model review of your diff |
| Automated PR reviews | CLI review git | Multi-model review in GitHub Actions |
| Planning a refactor | review_plan | Catch design issues before coding |
| Reviewing React components | review_frontend | Accessibility + performance + UX focus |
| Securing an API endpoint | review_backend | Security + architecture analysis |
| Want deeper discussion | discuss_with_council | Multi-turn conversation with context |
| Audit entire codebase | tps_audit | Flow, waste, bottlenecks analysis |> Full tool reference: See Tools Reference for all parameters and examples.
Reading the Results
Code Council shows confidence levels for each finding:
| Level | Meaning | Action |
|-------|---------|--------|
| Unanimous | All models agree | High confidence - fix this |
| Majority | Most models agree | Likely valid - investigate |
| Disagreement | Models conflict | Your judgment needed |
| Single | One model found this | Worth checking |
Configuration
Code Council works out of the box with sensible defaults. For customization:
- Configuration Guide - MCP client setup, config files, environment variables
- Model Selection - Choose models, pricing, performance tradeoffs
- Tools Reference - Detailed tool parameters and examples
$3
`json
{
"env": {
"OPENROUTER_API_KEY": "your-api-key",
"CODE_REVIEW_MODELS": ["anthropic/claude-sonnet-4.5", "openai/gpt-4o"]
}
}
`Cost
Default models are chosen for cost-effectiveness (~$0.01-0.05 per review).
Swap in Claude/GPT-4 for higher quality at higher cost (~$0.10-0.30 per review).
See Model Selection Guide for pricing details and optimization tips.
Requirements
- Node.js >= 18.0.0
- OpenRouter API key
- MCP-compatible client (Claude Desktop, Cursor, etc.)
Troubleshooting
"OPENROUTER_API_KEY environment variable is required"
Add the API key to the
Reviews are slow
This is expected when using multiple models. Consider using fewer models or faster models like Gemini Flash.
Models returning errors
Check your OpenRouter credits and model availability at status.openrouter.ai.
Contributions welcome! Please open an issue or PR.
MIT
- Documentation - Full docs and examples
- OpenRouter - Multi-model AI API
- Model Context Protocol - MCP specification
- Claude Desktop - MCP-compatible AI assistant