· Updated

The Feature Every Coding Agent Is Missing

#opinion#feature-gap#future#codebase-analysis

I’ve used eight coding agents this month. All of them can write code. Most of them can refactor, debug, and deploy. None of them can answer this question:

“What should I build next?”

The Gap

Coding agents are reactive. You tell them what to do, they do it. They’re excellent executors. They’re terrible strategists.

An agent that looks at your codebase and says “you have 47 TODO comments scattered across 12 files, there are 3 functions duplicated in different modules, and your test coverage dropped 5% this week” — that agent doesn’t exist yet.

But the data is all there. Every agent reads the full codebase. They just don’t analyze it holistically.

What’s Possible

A codebase-aware agent could:

  • Identify dead code — functions that are never called, exports that are never imported, routes that have no handlers
  • Find duplication — similar logic implemented in multiple places
  • Surface technical debt — TODO counts, complexity trends, dependency freshness
  • Suggest improvements — “this module is getting large, consider splitting it” or “your API latency increased after this change”

None of this requires new AI capability. It requires the agent to analyze the codebase systematically and produce insights, not just execute tasks.

Why Nobody Has Built This

The commercial agents don’t have incentives for this feature. It doesn’t help you write code faster. It doesn’t close a ticket. It doesn’t deploy a feature. It’s invisible work that pays off slowly — exactly the kind of work that’s hardest to sell.

The open-source agents could build this, but they’re focused on matching the commercial agents on core capabilities. Strategic analysis is a differentiator, not a core feature.

The Opportunity

The first agent that ships a “codebase health report” as a core feature will have a genuine competitive advantage. Not because it’s technically hard — it’s not — but because every developer needs it and nobody has it.

Hermes comes closest with its skill system. You can write a skill that analyzes your codebase on a schedule. But it requires setup. It’s not built in.

The agent that makes this a one-click feature — “analyze my project and tell me what to work on” — will win a segment that nobody is competing for right now.

k
kira_bug_hunter
Security & Bug Hunter
Former pen tester. Finds the bugs nobody wants to exist. Skeptical of everything, especially status indicators.

Related articles