PR-Agent

PR-Agent is an open-source, AI-powered code review agent that operates on pull requests. Its README styles it “The Original Open-Source PR Reviewer” and describes it as a community-maintained legacy project of Qodo, the company that built it. It is published under an MIT license, with its own site at https://www.pr-agent.ai and documentation at https://docs.pr-agent.ai.

The project’s functions are exposed as slash commands, run either as a comment on a pull request or from a CLI against a pull request URL. /describe, /review, /improve and /ask act on a pull request; /similar_issue is scoped to an issue instead. The README states that each of /review, /improve and /ask runs in a single LLM call, which it puts at roughly thirty seconds and a low cost.

The README distinguishes the project from Qodo’s own current offering, which it describes as a feature-rich, context-aware experience, while labelling PR-Agent itself a community-maintained legacy project. It is explicit that this repository is not Qodo’s free tier for open-source projects, which is a separate thing Qodo offers.

Capabilities

The commands above are the interface. The README glosses only two of them — /ask takes a free-text question about the pull request, and /similar_issue finds issues in the repository resembling a given one — and lists the rest bare, deferring to its documentation site for the feature and git provider support matrix and for each tool’s own page. It does note that /help_docs has been temporarily disabled since v0.36.1 pending a fix for a credential-exposure issue, so the command set a given installation exposes depends on which version it is running.

Two properties are claimed for the project as a whole rather than for any one command. A PR compression strategy is credited with letting it process both small and large pull requests effectively. Separately, prompting is JSON-based, which the README gives as what makes review categories and behaviour customizable through configuration files rather than by editing the tool.

Nothing in the README ties the review output to a particular model. Models named as usable are OpenAI’s GPT, Anthropic’s Claude, Google’s Gemini, DeepSeek and Mistral, and beyond those any model reachable through LiteLLM — a list that the README extends to Azure OpenAI, AWS Bedrock, Vertex AI, Databricks, OpenRouter and Ollama.

Adoption & Ecosystem

PR-Agent is platform-agnostic across both the forge it reviews on and the way it is deployed. The git providers it supports are GitHub, GitLab, BitBucket, Azure DevOps and Gitea; the deployment shapes are a CLI, a GitHub Action, Docker, a self-hosted instance and webhooks. The quick-start paths the README leads with are a GitHub Actions workflow referencing the project’s own action, and pip install pr-agent followed by a CLI invocation carrying an API key in the environment.

Docker images moved namespace partway through the project’s life: releases from 0.34.2 onward are published under pragent/pr-agent, while releases up to and including v0.31 remain at the earlier codiumai/pr-agent namespace as a frozen archive that receives no new images. The README asks that pinned image references be updated when upgrading across that boundary.

The project’s governance has moved as well. Qodo donated it to the open-source community; it now lives in its own GitHub organization — the repository is published as The-PR-Agent/pr-agent, and the quick-start GitHub Actions workflow references the action as the-pr-agent/pr-agent@main — is described as fully community-owned and open to additional maintainers, and has gained its first external maintainer. The README states that it is in the process of being donated to an open-source foundation, and that its documentation has moved to docs.pr-agent.ai. Qodo remains its gold sponsor, under a heading stating that the project’s ongoing development is supported by its sponsors.