About ArchSetu
Built by developers, for developers
ArchSetu started as a VS Code extension to help developers navigate large, unfamiliar codebases. Understanding architecture should not require reading ten thousand lines of code. So we built the tool we always wished existed.
The story
Every developer has faced this: you join a new team, inherit a legacy codebase, or want to contribute to an open-source project. The first question is always the same: where do I even start?
We spent years watching developers burn hours on this problem. Reading file by file. Running grep sessions. Asking senior engineers who themselves were guessing. The mental model of how code connects is almost never documented, because it changes every sprint.
ArchSetu solves this with pure static analysis. We parse code the same way a compiler does, by reading structure rather than executing logic. No AI interpreting things, no cloud model guessing at what your function does. Just the actual call graph, dead code, and complexity score.
We built ArchSetu in India, because Indian engineering teams face this problem acutely. Massive legacy codebases, fast-growing startups with minimal documentation, and remote teams that cannot rely on hallway conversations to share architecture knowledge.
Why this matters more now, not less
AI writes code faster than any team can review it
A growing share of the code in any real codebase was generated by an assistant, not written and remembered by a person. That is not a problem AI review tools can fully solve either - asking one language model to grade code another language model wrote is still a guess, just a more expensive one. Architecture understanding has to come from the code itself, computed the same way every time. That is the actual gap ArchSetu is built to close, and it gets more relevant every quarter code generation accelerates, not less.
A real parser per language, not one regex pretending to understand 47 of them
Every supported language gets its own syntax-tree parser, built to that language's actual grammar. A Python decorator and a Go struct tag are not the same shape - ArchSetu never treats them like they are.
A real call graph, traced, not guessed
Every function, every caller, every import resolved from the syntax tree - the same structural information a compiler needs, not a language model's best guess at what a function "probably" does based on its name.
The same repo produces the same report, every time
No sampling, no temperature, no "run it twice, get two different answers." A health score computed today is the same score computed next week on the same commit - you can build a CI gate on top of that. You cannot build one on an LLM's mood.
Security findings, disclosed responsibly by design
A finding's exact file and line is only ever shown to that repository's verified owner, or the account that submitted a private analysis. Everyone else sees a severity count, not a map of what to exploit. That is enforced in the product, not a policy on a page nobody reads.
Your source code is never retained
A repository is cloned into an isolated environment, analyzed, and the clone is deleted immediately after. There is no step in this pipeline where your code is used to train anything, sold, or kept around "just in case."
Repositories analysed
1,070
Completed runs
1,373
Languages in the wild
23/ 47
Of the languages the engine parses, this many have turned up in a repository somebody actually ran.
Functions, one run
768K
torvalds/linux, in a single pass.
Live counts, read from ArchSetu’s own database on each rebuild.
What we stand for
Static Analysis Only
No AI. No LLMs. No hallucinations. Every insight is derived from parsing your actual code structure.
Privacy by Design
We clone your repo, analyze it in an isolated container, then delete it immediately. We never store source code.
Speed Over Everything
Full repo analysis in under 60 seconds. Results are cached and shareable instantly.
Built for Teams
From solo developers to engineering teams enforcing architecture rules in CI. ArchSetu scales with you.
Where this interface comes from
This site is assembled out of a small vocabulary of Indian architectural elements, each one chosen because the building solved the same problem the interface has. These are the sources.
तोरण torana
Page and section openers
The Sanchi toranas, Madhya Pradesh
1st century BCE
A gateway stands free of any wall and encloses nothing - it marks that you are crossing into somewhere. So here it opens a page, once, and nothing else.
बावली baoli
Nested drill-downs
Rani ki Vav, Patan, Gujarat
11th century, commissioned by Udayamati
A stepwell descends in flights, and every flight is also a floor you can stop on. That is a drill-down with landings, which is what a nested result list is.
जाली jaali
Empty, loading and withheld states
The Sidi Saiyyed mosque screens, Ahmedabad, and Laurie Baker’s perforated brickwork
1572 / 20th century
A pierced screen is calculated, not decorated: the size of the perforation is the setting. It marks a surface where something is absent, pending, or deliberately held back.
The calibrated scale, and the measuring idea behind the product
Jantar Mantar, Jaipur and Delhi
1724–34, Sawai Jai Singh II
Masonry instruments that measure by casting a shadow across a graduated stone arc. No lens, no interpretation, the same reading twice - which is what a deterministic analysis engine claims to be.
The stepped reading order of the dashboard panels
Sangath, Ahmedabad
1981, B. V. Doshi
A section that steps down in vaults, so the building is read in order rather than taken in all at once.
Nine columns instead of twelve
Jawahar Kala Kendra, Jaipur
1992, Charles Correa
Planned as nine squares, one per graha, on the same mandala Jai Singh laid Jaipur out on. The dashboard grid is nine squares for that reason, not because a CSS framework shipped twelve.
These are references, not endorsements. None of the architects, their estates, or the institutions that hold these buildings is associated with ArchSetu, has reviewed this site, or has any connection to this company. They are named here because the design borrowed from their work, and the borrowing should be visible.
Proudly made in India
ArchSetu is designed, built, and maintained in India. We believe world-class developer tools can be built from India, for the world. Our infrastructure is hosted on global cloud providers to ensure speed everywhere, but our heart and team are in India.
Built with love from India