AI Code Review
Agent
An open source AI agent that reviews every PR around the clock. No API keys. No SaaS lock-in. Runs on its own VM.
Code review is the bottleneck. AI-generated code is flooding pull requests faster than your team can read it, and the tools that promise to help either lock you into a SaaS subscription or require API keys from OpenAI or Anthropic.
LiberClaw lets you deploy an open source AI code review agent that runs on its own VM, uses open models, and reviews every PR around the clock. Set up in under 5 minutes.
How It Works
Create an Agent
Sign in at app.liberclaw.ai and create a new agent. Give it a name like "PR Reviewer" and connect it to your GitHub repository.
Define the Review Skill
Write a SKILL.md file that tells the agent what to do: read the PR diff, check for bugs and logic errors, flag missing tests, and post a review comment. Plain markdown, no code required.
Deploy
Click deploy. LiberClaw assigns a dedicated virtual machine from a pre-warmed pool, installs your agent, and brings it online with its own HTTPS endpoint. Most deployments finish in under a minute.
It Reviews PRs
The agent runs on a heartbeat. It checks for new pull requests on a schedule, reads the diff, reasons about the changes, and posts review comments directly on GitHub. It works at 3am, on weekends, and on holidays.
Why LiberClaw
The entire platform is on GitHub. The agent code, the deployment logic, and the platform itself are all open. You can read every line, fork it, or run the whole stack yourself.
Each agent gets a dedicated virtual machine with its own filesystem, database, and HTTPS endpoint. A code review agent that processes sensitive diffs is completely isolated from every other agent on the network.
Inference runs through LibertAI using open models: Qwen3 Coder for code analysis and GLM-4.7 for general reasoning. No usage tracking by third parties. No surprise bills.
Once deployed, your agent runs around the clock and builds context about your codebase over time. The PR that lands at 3am on a Saturday gets the same review quality as the one submitted at 2pm on a Tuesday.
Pre-warmed virtual machines mean no cold starts. Create an agent, paste your skill, deploy. The agent is live and reviewing PRs in minutes, not hours.
How LiberClaw Compares
| LiberClaw | CodeRabbit | Qodo / PR-Agent | Manual Review | |
|---|---|---|---|---|
| Open source | check_circle Fully | Partially | check_circle Yes | N/A |
| Self-hosted | check_circle Own VM | close SaaS only | Self-managed | N/A |
| No API keys | check_circle | close | close BYOK | N/A |
| Runs 24/7 | check_circle Managed VM | check_circle SaaS | Your infra | close |
| Setup time | Under 5 min | Minutes | Hours | N/A |
| VM isolation | check_circle | close | close | N/A |
| Persistent memory | check_circle | close | close | Human memory |
CodeRabbit works well if you are comfortable with a SaaS model and proprietary inference. If you are looking for an open source alternative, Qodo's PR-Agent is genuinely self-hostable, but you need to bring your own API keys and manage the infrastructure yourself. LiberClaw sits in the middle: open source with managed deployment, open-model inference, and no API keys to configure.
Frequently Asked Questions
How do I automate code review with AI?
Deploy an AI pull request reviewer that watches your repository for new PRs. With LiberClaw, you create an agent, define a review skill in a markdown file, and deploy it to a dedicated VM. The agent checks for new PRs on a schedule, reads the diff, and posts review comments on GitHub. The entire setup takes under 5 minutes, and you do not need an API key from any provider.
What is the best open source AI code review tool?
It depends on what you need. Qodo's PR-Agent is a solid open source option if you want to bring your own API keys and manage your own infrastructure. LiberClaw is a good fit if you want an open source AI code review agent that runs on a managed VM with open-model inference and no API keys. CodeRabbit is strong but not fully open source.
Can I run AI code review without an API key?
Yes. LiberClaw agents use LibertAI for inference, which runs open models (Qwen3 Coder, GLM-4.7) on decentralized infrastructure. You can deploy a self-hosted AI code review agent without configuring API keys from OpenAI, Anthropic, or any other provider. The inference is included.
Explore More
Deploy Your Code Review Agent
Free tier includes 2 agents. No credit card required.
Get Started Free arrow_forward