Products & Tools

Qodo Automates the Merge Decision With Policy-Driven PR Approvals

Qodo's November 18 release adds beta Auto Approval Workflows and an Agent Session Review preview, moving the merge decision itself into the AI code toolchain.

Qodo Adds Policy-Driven Approvals and Session Reviews for AI-Generated Code
Qodo Adds Policy-Driven Approvals and Session Reviews for AI-Generated CodePexels / pixabay
By Sophie Lindqvist5 min read

Why it matters

  • Qodo announced Auto Approval Workflows (beta) and Agent Session Review (research preview) on November 18.
  • Qodo's bot approval counts as one approval; required reviewers and code-owner rules still apply.
  • Each new commit triggers reassessment, and approval is withdrawn if the updated PR no longer qualifies.
  • Agent Session Review consolidates multi-repository PRs from one coding session into a single assessment.
  • Workflows can apply at workspace, Git organization, or individual repository level.

Qodo shipped two governance features for AI-generated code on November 18: Auto Approval Workflows, available in beta, and Agent Session Review, a research preview that consolidates multi-repository agent work into a single assessment.

The release targets a specific point in the development pipeline that most AI code tools skip past. Reviewing a pull request finds problems; approving it records a decision that the change may move forward under the team's process. Qodo's argument is that automating the first step does not clear the queue at the second. Routine pull requests can still sit waiting for a human approval long after automated review finishes.

The announcement matters because the industry's bottleneck is shifting. As coding agents produce more changes faster, the scarce resource is no longer code generation — it's the judgment about which code earns permission to proceed. Qodo is one of the first vendors to build that decision process directly into the toolchain.

What does Auto Approval Workflows actually do?

According to Qodo's announcement, engineering teams can set rules that define which changes qualify for automatic handling. When Qodo finishes reviewing a pull request, a matching workflow can approve it or apply a "Qodo approved" label. Teams can start from templates designed for low-risk work and apply workflows at three levels:

  • the individual repository
  • the Git organization
  • the workspace

The distinction between a label and an approval is the load-bearing detail here. A label is a signal; it does not authorize a merge on its own. Qodo says its bot's approval counts as one approval, while existing required reviewers and code-owner requirements still apply. In other words, this is not unrestricted automatic deployment — the human gates that a team configured before Qodo arrived stay in place.

That design moves a repeatable decision from people into policy. Instead of an engineer rediscovering the same acceptance criteria for every routine update, the team writes those criteria once, as a workflow. Qodo's framing implies the honest caveat: the value depends on whether the policy captures the conditions that genuinely make a change acceptable, not on how many approvals the system can issue. A bad policy automates bad judgments faster.

The takeaway: Qodo is automating the merge decision only for changes a team can describe in rules — and leaving everything else to humans.

How does the system keep human review around critical code?

Qodo's guards let teams mark repositories and authors that should never receive automatic approval. That gives organizations an explicit boundary: delegate a narrow class of routine decisions to policy, retain direct review everywhere else. The announcement positions this as more specific than asking engineers to trust all AI-generated code, and it keeps ownership of the policy with the organization deploying the tool.

One mechanism does the heavy lifting for correctness: each new commit triggers a reassessment. If the updated pull request no longer qualifies under the workflow, Qodo withdraws the approval. A PR Triage view records the approval status of each change and shows which workflow or guard produced the decision.

The reassessment matters more than it might sound. An approval is a judgment about one particular version of the code. A small fix can grow into a larger change as commits pile up. If the approval stayed attached regardless, it would describe an early snapshot rather than the final implementation — and say less about the shipped code than it appeared to.

The takeaway: approvals in this system are conditional and revocable, which is what separates a policy from a rubber stamp.

What is Agent Session Review for?

The second feature tackles the opposite problem. Coding agents working on larger features can produce several apparently reasonable changes spread across multiple repositories, and each pull request can look fine in isolation while the whole feature doesn't hold together. A backend change and a client update, for example, may rest on different assumptions about how they'll interact.

Agent Session Review, running through Qodo's Agentic Toolbox, gathers the PRs produced during a coding session — including multi-repository work — into one assessment. Qodo says the result connects:

  • implementation decisions
  • deviations from the original plan
  • merge order
  • risks, tied to evidence in the codebase

For an engineer, the useful unit of review here is the behavior of a complete feature, not the boundaries of any single repository. The session-level view is meant to make those cross-repository relationships inspectable rather than something you reconstruct by hand from a stack of PRs.

The research-preview label is doing real work, and Qodo's announcement treats it that way. The feature describes a new way to organize evidence for review; it does not demonstrate that every unintended behavior will be caught. Engineers still have to decide whether the implementation matches the intended requirements.

The takeaway: session review changes the granularity of review — from "is this PR acceptable?" to "is this feature, as the agent built it, what we asked for?"

Why this matters beyond Qodo

Taken together, the two features sketch a practical division of labor between policy and people: workflows handle changes whose acceptance conditions can be stated clearly, while engineers examine the decisions and interactions that require broader judgment. That's a different proposition from the industry's louder claims about autonomous coding, and it's deliberately modest.

It's also worth noting what the announcement does not contain. Qodo provides no independent evidence of faster delivery or fewer defects, and the company's own framing suggests those outcomes should not be assumed from automated approval alone. The beta and research-preview designations carry the same message.

The larger shift Qodo is betting on is measurable: development teams are moving from measuring how quickly an agent produces code to understanding how that code earns permission to proceed. If that shift continues, approval policy and session-level review will sit at the center of the AI development toolchain — and Qodo now has an early position there.

Source: unite.ai

Share this article:

More from Sophie Lindqvist

Sophie Lindqvist

Show full bio

Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.

777 articles

Related articles

  1. Qodo plugs organizational context into OpenAI Codex for AI code review
  2. 83% of Firms Run Unapproved AI Coding Agents, BlueFlag Research Finds
  3. Google now runs six automated gates, including an AI reviewer, on every agent skill
  4. CodeRabbit says Codex helped it ship 20x more software
  5. New Relic's Autopilot Kubernetes Agent Hits Public Preview

« Previous articleNext article »