Agentic coding startup Baz Technologies Inc. said today it’s launching a new platform that sits between developers and the code bases they’re working on in order to catch software vulnerabilities before they enter production workflows.

The launch came as the company clinched $9 million in extended seed funding, bringing its total amount raised so far to $17 million. The latest round was co-led by existing investors Battery Ventures and Boldstart Ventures, and saw participation from new backers including AFG Partners and Disruptive VC.

But the bigger news was the launch of Baz Planner, which debuted today at the AI Engineer World’s Fair event taking place in San Francisco today. Baz Planner is a new gateway that automatically routes every new idea through dynamic loops that can instantly detect and understand the root cause of new vulnerabilities, then proactively rewrite the coding plan to eliminate them.

The tool follows in the footsteps of its popular AI Code Review tool that launched last year, and currently ranks in first place on the precision-weighted Code Review Bench. With that tool, development teams can govern and secure their AI code with AI agents designed to enforce coding standards across product, design, architecture, security and reliability issues.

Though cyber-capable AI models that can autonomously scan millions of lines of code to discover subtle flaws are nothing new, Baz Planner is designed to step code security up a notch by scrutinizing every ad-hoc change made to a codebase against both the current and prospective new architecture. Any problematic changes introduced will be flagged before the new code is even saved.

The startup was founded by a team of former Palo Alto Networks Inc. engineers who helped scale that company’s cloud application security business to become one of the leading code-to-cloud security firms. The team is bringing the same four principles that shaped cloud development to AI-generated code: It must be observable, explainable, predictable and reproducible, so as to avoid flaws creeping in.

With Baz Planner, every new suggestion made by an AI model is evaluated against a strict risk matrix that blocks unsafe paths and enforces defined boundaries. It will only allow the code to be shipped into production after applying rigorous risk mitigation. In this way, it has helped early adopters to reduce downstream rework by more than 65%, as measured by the frequency of reverts and hotfixes that have to be done following a merge.

Baz co-founder and Chief Executive Guy Eisenkot said customers have been pushing the company to go beyond code reviews and intervene much earlier, during the planning stages. “That’s where bugs and vulnerabilities are cheapest to eliminate,” he explained. “Baz exists because they refuse to accept that AI-generated code means blindly accepting risk.”

Rather than focusing on style and syntax, Baz Planner analyzes how new code will impact runtime to try and catch bugs, silent regressions and other security flaws. The tool employs four specialized agents that work together to review every new code suggestion, including a spec reviewer agent that validates new code against product requirements, designs and expected behavior.

Then, the advanced security agent reasons across authorization and network boundaries, infrastructure, pipelines and finally the application code to uncover any new vulnerabilities. The site reliability engineer agent will correlate repository changes with production telemetry to identify risks pertaining to performance, reliability and observability. Finally the fixer agent applies and validates every code change that’s been determined as “safe” in an isolated runtime environment.

“Guy and the Baz team built the code-to-cloud security playbook at Palo Alto Networks and they are now applying that same rigor to AI-native engineering,” said Battery Ventures Partner Barak Schoster. “As development teams deploy fleets of coding agents, Baz is becoming the super harness that coordinates them, from spec-driven development and UI review to security, quality, reliability and planning to ensure AI-generated code ships safely at scale.”

Image: Baz Technologies