Does VizDev Use AI? Working With and Without Coding Agents

In the current technology climate, "Is VizDev AI?" is one of our most asked questions, so we'll answer it upfront: VizDev is not an AI tool. VizDev makes AI tools better.

Clearly there's more nuance in this topic than fits in the top line, so we wanted to take the time to explain more and talk about why VizDev is the way it is. Because ultimately how we chose to build VizDev reflects our values: building tools that empower humans to do more, and doing things the right way.

No model makes your map

VizDev doesn't use LLMs to map your code; the local analysis engine is written in Rust and the visualizations are in TypeScript. We do this because it makes good engineering sense: our analysis is blazing fast, deterministic, and portable. Our visualizations turn the facts from that analysis into intuitive workflows and insights about how code works.

Our team's experience with program analysis meant we knew we could build the tools to extract the answers we needed from code. While it might not be the easiest way to get a solution, we think it's the right way. This means the analysis focuses on just the ground truth of the code itself. Everything is linked to source locations, not just to help navigation, but to make it so you can always look at the code and verify anything.

The second important consequence to designing things this way is that VizDev just works out of the box, without requiring AI or even a network connection. All of the analysis is performed locally on your machine so you maintain control of your code and the analysis. You don't have to worry where your code goes, what's going to happen to your next token bill, or if an upstream service outage could ruin your day.

This is why we say VizDev isn't AI. But it's not the whole story, because we built it with users in mind; prioritizing exportability, interoperability, and capability. And AI can bring some impressive capabilities.

Where AI comes in

VizDev isn't built to replace agentic tools like Claude Code, it's built to augment them. VizDev provides a knowledge graph and data model that makes agents better at understanding your project, and visualizations to make their insights more usable for you. So while VizDev is built for humans, agents really get a lot out of it too.

How do we do this? We use integrations like Agent Skills and MCP to allow agents to use the capabilities of VizDev and turn it into a superpower for agents. Focusing on integrations also make VizDev agent and harness agnostic, so it just works for Codex, Cursor, OpenCode, and so on.

We also use agents to accomplish tasks that would normally require flexibility or a lot of user interaction. The results we've seen so far are impressive because it allows the flexibility and power of these agents to be reflected in more intuitive workflows and processes. Analysis and data wrangling that used to be a pain to set up is now possible by asking an agent a simple question, and the result is still grounded in our visuals and thus in the code.

We built VizDev to help humans understand code, and this led to VizDev becoming a way to help humans not just keep up with AI, but a way for them to become more knowledgeable and engaged as they work with coding agents. Our visualizations are geared toward how the human brain works and the workflows that people actually use. Incidentally, this helps people get the most out of AI tools.

VizDev vs. asking your agent to explain the codebase

VizDev initial design goals were to make code review easier and to make code onboarding easier. But as it's become more common to use agents for all kinds of tasks, certain questions have become more common.

"Why not just ask an agent to explain the code?"

If you're like us and have spent some time with coding agents, you might have experienced the dreaded menace of TEXT FATIGUE. It's not just you, and it's not a personal weakness. The human brain does not efficiently process walls of text, and after a few hours of reading endlessly scrolling words, it's a breath of fresh air to be able to explore an interactive visual instead.

"Ok, but why not just ask an agent to make all your diagrams for you?"

Sit with that a minute and the answer becomes clear: that approach is slower, non-deterministic, costs tokens, and is separate from the code you want to know about. You can go that road, but there's a better way.

Our model was custom-built by a team with deep expertise in code security, and the express goal was enabling human understanding. The model wasn't built to make IDEs better, it was built to explain code. It works great for both humans and agents, which helps make the difficult problems of working with code easier and more intuitive.

VizDev tackles real engineering team concerns

With software development teams moving faster than ever, VizDev is one of the few tools focused on helping people keep up by being built to make humans more effective and efficient. We focus on speed, transparency, and interoperability... because we know how great it is to work with tools that give you that!

The advantages VizDev provides also help with some real fundamental issues that teams deal with. By giving people interactive and explorable visualizations, we increase engagement and contribute to a more enjoyable developer experience. By using VizDev to read the structure of your codebase, we can reduce token spend by just giving answers to agents directly. By making it more intuitive to understand big chunks of code, we help people move through common problems more quickly and confidently. We're just starting to measure these results more directly, and look forward to sharing more about the impact we've seen.

We also focus on the problems we hear from users, in organizations ranging from cybersecurity teams to software shops run by unconventional builders. With so many ideas about tools and workflows being released all the time, we find it helpful to work on the core problems that it seems everyone has issues with. As a result, we're continually rolling out more capability and features, but the core experience delivers immense value because code understanding is the root of most code problems.

VizDev is built to help humans, and their agents

VizDev is built for people, and so we focus on making it awesome for all our users, delivering fast and intuitive workflows for everyone, and slick integrations for users with coding agents.

It's not a question of either/or, it's a both/and. People need visuals, agents want knowledge graphs. VizDev delivers both. This means that our momentum is increasing, more is becoming possible as we build, and our users reap the benefit.

We still have room for a small number of design partners, so we're looking for indie developers, engineering teams, and AppSec professionals. People willing to try VizDev on their real codebases, with their real workflows. We need partners who will use it in ways we didn't anticipate and tell us what's working and what isn't. Join the beta and help us chart what comes next.