Sourcery

Sourcery has moved from an editor refactoring coach to an AI code-review and security platform that meets a change wherever it happens: on pull requests in GitHub or GitLab, across repositories, inside the IDE before you commit, and with agents that fix what the reviewer flags. Real-time feedback in the editor is still in the pitch, with one-click fixes, but the centre of gravity is automated review — summarised PRs, architecture diagrams, custom team guidelines and vulnerability scanning at configurable frequency, plus an Enterprise Security tier with SOC 2 attestation. Pricing reads Open Source (free on public projects), Pro at $12 per seat a month, Team at $24 and Enterprise by quotation, with annual billing roughly 20% lower.

Where it loses: this is not a generator that writes your feature, and if you want an agent to implement tickets you are at the wrong door — the value here is the gate, not the drafting. Where it wins over general-purpose reviewers is that quality advice and security findings arrive in one pass and land in tools your developers already look at. Check three things against your own repository before adopting: language coverage (the marketing names surfaces and editors — VS Code and PyCharm appear in the site footer — rather than publishing a language list), what the per-plan repository counts and scan-frequency caps mean for a large monorepo, and whether Team buys the review depth you need or the security tier is the real reason to talk to sales.