Qodo

🔗 Try Qodo for free → https://qodo.ai Jonathan and Shlomo, Qodo's Field CTO, walk through the Qodo's Agentic Toolbox: a set of skills, a CLI, and an MCP server that bring Qodo's full review engine and codebase context directly into your coding agent, before a pull request ever exists. The starting point is a shift most engineering teams are living through. AI taking the first pass at code review was good enough a year ago. Now that agents are generating most of the code, the PR is no longer the right place to catch problems. The session makes the case for running the exact same review locally, inside the agent harness, so the PR opens in a state that is already close to mergeable. They also cover the wider arc: triaging PRs so humans only look at the ones that matter, the L1 to L5 spectrum of semi-autonomous software factories, and why trust in code is falling even as productivity rises. Plus a practical adoption order, a live install and demo, and customer examples of the CLI running inside GitHub Actions. [00:00] Introduction [01:03] Why AI taking the first pass at review stopped being enough [03:04] Shifting review left, before the PR exists [04:34] Software factories and the L1 to L5 autonomy spectrum [06:46] Productivity is up, trust in code is at an all-time low [08:01] What Qodo learns on day one: org rules, cross-repo dependencies, PR history [11:23] How far back Qodo reads your historical PRs [13:21] Inside the Agentic Toolbox: skills, CLI, and MCP [14:54] Local review runs the same engine as the PR review [16:11] Codebase Wisdom and agentic search without cloning every repo [19:38] Auto-approving, blocking, and labeling PRs [22:22] Where to start: local review, then planning, then the factory [23:16] Deep findings for agents, low noise for humans [30:43] Fast, balanced, and deep review modes inside a loop [33:42] Demo: installing and running the Agentic Toolbox [38:01] Reading a local review's findings [41:44] CLI vs MCP, and running Qodo inside GitHub Actions 💡 Key takeaways: → The PR is becoming the final gate, not the first one. When agents write the code, the review needs to happen inside the agent loop, and the PR becomes a confirmation step rather than a discovery step → Humans and agents need different reviews. Agents can absorb every finding, so local review should go as deep as possible. Human-facing PR review should be tuned for low noise so people only see what actually matters → Developers fix more than blockers when the review happens locally. Once code reaches a PR, teams tend to skip the reliability, maintainability, and defensive coding improvements. Catch them in the IDE and they usually get fixed in the same pass → Local review is not a prompt wrapped in a skill. It calls Qodo's backend and runs the full review, with cross-repo dependency mapping, org rules, and learning from a year of historical PRs, so your agent gets organizational context, not just repository context → Codebase Wisdom lets an agent search repos it has never cloned. For orgs with thousands of repos, that means your agent can pull real cross-repo context and past bugs into a plan before a single line is written → Verification is what keeps a software factory standing up. Agents without a review loop produce garbage at scale, which is why Qodo's internal factory started favoring the Agentic Toolbox on its own About Qodo Qodo is the AI code review platform built for engineering teams that ship at scale. As AI-generated code becomes the norm, Qodo provides the quality layer between "AI wrote it" and "production-ready", analyzing full codebase context, enforcing team-specific standards through a living rules system, and catching critical issues, breaking changes, and duplicated logic before they merge. With 2M+ installations and 4M+ PRs reviewed every year, Qodo gives engineering teams the precision and confidence to accelerate without compromising quality. Start free at qodo.ai. #AgenticToolbox #AICodeReview #CodeQuality #SoftwareEngineering #DeveloperTools #CodingAgents #SoftwareFactory #ShiftLeft #PullRequests #ClaudeCode #DevTools #AIEngineering