Qodo

Qodo is the reviewer you want when the problem is governance rather than typos: your team writes the rules down, the bot applies them to every pull request on every repo, and a lead can see which rules keep failing. The trait that decides it is that the rules are enforced as written, so comments stop being one reviewer's taste and start being policy, which is the only way review scales past ten engineers.

What you will not love is a bill that takes a spreadsheet to predict. Pick it for a team with written standards and someone who reads dashboards. Pick CodeRabbit if you are solo.

The pricing page lists a $30 seat and credits at $0.012 each, and it does not list credits per review, so the second meter has no published exchange rate into the unit you actually consume. There is a 14-day trial instead of a free tier, enough time to see the comments and not the invoice.

The failure mode is financial: a busy repository with many small pull requests burns credits in proportion to activity, and nothing in the docs lets you estimate it before signing. What it does right: one reviewer across GitHub, GitLab, Bitbucket and Azure DevOps, so a mixed shop gets one policy instead of four.

Qodo's Agentic Toolbox is documented: review, rules and codebase-context tools are packaged as a CLI, as skills for Claude Code, Codex and Kiro, and as an MCP server, so the reviewer becomes a callable component another agent can invoke before a pull request exists. That is a principled shape; it moves review to edit time.

The central claim, that it reviews well, has no number attached: no benchmark appears anywhere on the site, and no methodology could be audited if one did. Governance features survive model changes because they are rules; unbenchmarked review quality does not, because it is the model. The observation: the documented part is the plumbing.

The rename from CodiumAI to Qodo in September 2024 is one story told under a new name: a developer tool became a platform sold to engineering leadership, where the buyer signs the invoice and the developer reads the comment, and pricing power lives with the signer. That is the correct customer, and a crowded one, because every review bot on this board is making the same turn.

The platforms it plugs into are also its natural acquirers, Atlassian and GitLab first, since a governance dashboard is a feature they would rather own than compete with. Position: reasonable at team scale, keep the rules exportable in case the logo changes.

The demo is a review bot with a dashboard. Sixty seats at $30 is $1,800 a month before the credit packs, and credit consumption per review is the number nobody budgets until the second invoice, so finance needs a pilot repository and a month of data. SSO and on-prem are Enterprise-only at custom pricing, which means the security questionnaire waits for a sales call.

CI fit is native, since the product lives on the pull request. Onboarding is a rules file and an afternoon. Approved with conditions: Enterprise quote for SSO, credit burn measured on one repo, and a hard cap on the packs.

PR-Agent was the open-source thing; now it is community-maintained under a separate organization, which is open source with the good part behind a login. Bring-your-own-key is Enterprise-only, so the trial cannot use my key, and there are no local models at all; my code goes to their cloud or nowhere.

What I grant, grudgingly, is the Agentic Toolbox: curl -fsSL https://get.qodo.ai | sh gets me a CLI and an MCP server my own agent can call from its config, so the reviewer at least speaks a protocol I control. A fork is impossible; the parts worth forking already left the building.