Octopus vs Revue

Octopus indexes a codebase and uses that context to review pull requests automatically. When a pull request opens, it retrieves related indexed code through vector search and sends the changes and context to a configured language model. Findings receive severity ratings and appear as inline comments. The GitHub integration also reviews updates, posts check runs, and can create issues from findings. Octopus supports GitHub, GitLab, Bitbucket, and Forgejo, and includes source-cited codebase chat plus a CLI for reviews, context queries, and repository checks. Teams can use knowledge-base documents to guide reviews, with rules, repositories, and reviewer settings shared across an organization. AI options include providers such as Anthropic Claude, OpenAI, Google, xAI Grok, Alibaba Qwen, Cohere, and OpenRouter. It can be self-hosted with Docker Compose, PostgreSQL, and Qdrant, with optional local models and embeddings through Ollama. The cloud service stores review and indexing data; findings persist through the PR lifetime and 90 days after closure, while indexed embeddings remain as long as a repository is connected. Free self-hosted and cloud tiers are listed. Live team telemetry requires a paid organization.

Revue is a multi-agent AI code reviewer for local changes and pull requests. Its Claude Code skill reviews staged changes, while CI mode can review pull requests or merge requests on GitHub, GitLab, and Bitbucket and post findings as inline comments. Six specialist agents cover security, performance, code quality, architecture, synthesis, and verification. Revue says it checks findings against the code and anchors them to relevant lines. Users supply their own AI provider API key; listed providers include OpenRouter, Anthropic, OpenAI, Azure OpenAI, and custom OpenAI-compatible endpoints. For local reviews, code runs in the user’s environment and source code is not received, stored, or processed by Revue’s servers. The service may log code-change metadata, review counts, timestamps, repository and platform metadata, and diagnostic data. The default diff limit is 2,000 lines and can be raised in.revue.yml. Windows support requires WSL2; there is no native Windows build, and Windows on ARM is unsupported. Free costs nothing, with unlimited reviews for 30 days and 10 per month afterward. Indie is $9 a month; Pro is $29 a month.