mrkeyoor.com_
Tue 01 Sept 16:08 UTC
Dev Toolsevaluationupdated 26 Aug 2026

archify review

Archify is an agent skill that turns a repository or system description into a polished, interactive architecture map. It is meant to replace hand-maintained diagrams with typed, checked artifacts that can be searched, presented, compared, and exported.

+21,790stars / 7d
Verdict

Our run installed Archify in 22 seconds, but 2 of 710 tests failed, so it is promising rather than friction-free. Use it if interactive, source-grounded maps would improve design reviews or onboarding and you can validate the generated artifact in your own environment. If all you need is a portable diagram embedded in Markdown, Mermaid or D2 is the simpler choice.

We ran it

Install✓ · 22s9 packages · 29 MB
Buildn/ano build script
Tests✗ · 130s708 passed · 2 failed of 710 (node:test)
Known vulns00 critical · 0 high · 0 moderate · 0 low (npm audit)
Repo439 files~37,512 lines of source · 38.3 MB · 3 CI workflows · tests dir

Answers from our run

Does archify build from source?

Dependencies installed in 22 seconds (9 packages), and the project has no separate build step. We cloned commit 8323a15 into a clean Debian container with 3 CPUs and no project-specific setup.

Do archify's tests pass?

Not all of them: 708 of 710 passed and 2 failed when we ran the project's own test command (node:test). Some failures need services or credentials a bare container does not have.

Does archify have known vulnerabilities in its dependencies?

npm audit found none in the dependency tree at the time of our run.

Who should not use archify?

Teams wanting automatic runtime discovery with no agent analysis or authored input

What are the alternatives to archify?

Mermaid, D2, Diagrams. Our run installed Archify in 22 seconds, but 2 of 710 tests failed, so it is promising rather than friction-free.

Setup4/522s install, but 2 of 710 tests failed
Docs5/5Detailed quick starts, scenarios, proofs, and export guidance
Community4/516,851 stars, fresh push, recent release, 21 open issues
Maturity3/5Strong validation design, but our suite was not fully green

Who it’s for

Engineering teams that need reviewable architecture maps tied to source
Developers using Raven, Cursor, Claude Code, Codex CLI, or OpenCode
Tech leads who want shareable diagrams for design reviews and onboarding
Teams willing to keep diagrams grounded in a typed JSON representation

Who it’s NOT for

Teams wanting automatic runtime discovery with no agent analysis or authored input
Buyers who require a fully green test suite before adoption
People who only need a quick static flowchart and would find the verification workflow excessive
Organizations that need a hosted multiuser architecture repository with permissions and governance

Setup reality

In our run, installation succeeded in 22 seconds, adding 9 packages and using 29 MB, and there was no build target to run. The test suite did not pass cleanly: node:test reported 708 passed and 2 failed out of 710 in 130 seconds. The final log lines mostly show successful or skipped visual-check cases, so they do not identify the cause of the 2 failures and we will not guess. The README makes first use look like one npx command, which is fair for installation, but a team evaluating reliability should budget time to reproduce and investigate those failures, especially because several real-browser checks are skipped unless ARCHIFY_CHROME is set.

It turns architecture documentation into a checked artifact

Archify addresses a familiar failure: architecture diagrams become stale because they are detached from the systems they describe. An agent analyzes a repository or bounded description, produces typed JSON, runs deterministic checks, and emits a self-contained HTML map. Version 2.15.0 is the latest tagged release, while the README calls 2.16.0-dev.0 the current development version.

The README names Raven, Cursor, Claude Code, Codex CLI, and OpenCode. Its best prompt asks for a deliberately bounded view, such as 8 to 12 core components, one primary path, dependencies, and trust boundaries. Archify pushes supporting detail into cards instead of multiplying edges.

Our run installed quickly, but the tests were not clean

We cloned commit 8323a15 into a fresh unprivileged Debian container with 3 CPUs and 8 GB of RAM. Installation succeeded in 22 seconds, added 9 packages, and occupied 29 MB. The checkout held 439 files, about 37,512 lines of source, and used 38.3 MB. There was no build script or target, so we skipped that step. Npm audit also reported 0 known vulnerabilities at every listed severity.

Our test run exited with status 1 after 130 seconds: node:test reported 708 passes and 2 failures out of 710. The supplied tail does not contain the two failing assertions, so it does not support a diagnosis. It does show browser checks skipped unless ARCHIFY_CHROME is configured, followed by passing cases for structured Chrome failures, overflow evidence, tiny node text, legend obstruction, and screenshot cleanup. The repository has 3 CI workflows and a tests directory, but no Dockerfile. The README's one-line install held up; full verification needs more investigation.

Grounded interaction is the reason to choose it

Archify offers 5 diagram types, 4 presets, light and dark themes, brand marks, and finite motion. The valuable features are behavioral: search nodes, inspect upstream or downstream authored reach, trace exact routes, compare semantic roles, and play guided stories. Its cautious distinction between authored relationships and runtime impact is good practice. A map can show what its source asserts without pretending that assertion came from production telemetry.

Change review is another concrete advantage. Archify compares 2 validated snapshots as Before, Delta, and After, recording added, removed, changed, moved, and rerouted facts. For a pull request that changes service boundaries or message flow, that could give reviewers a focused companion to the code diff. Named views and shareable URL states let one artifact support onboarding, incident discussion, and design review without forcing every audience into the same crowded canvas.

Export coverage is unusually broad. One checked source can produce self-contained HTML, PNG, SVG, WebM, and a 1200 by 630 share card. Route and reach cards preserve the wider diagram as context, reducing the chance that a highlighted path is mistaken for the whole system. The Proof Lab contains 11 checked-in scenarios with JSON sources, named views, and validation receipts. A separate case maps mco-org/mco at commit 9f1a1cf, giving evaluators an inspectable repository-derived example rather than only a product mockup.

Validation cannot prove that the analysis is true

Generated topology can still be wrong when repository analysis is incomplete, a prompt is vague, or the typed source encodes a mistaken relationship. Deterministic checks can prove internal consistency, but they cannot prove that a component omitted from the selected 8 to 12 nodes is irrelevant. Teams still need an owner to review boundaries, labels, and routes before accepting an export as an architecture record. The right model is assisted documentation with human review, not unattended truth generation.

There is also real complexity behind the viewer. Search, source links, stories, lenses, route probes, motion exports, and multiple diagram types create more surface area than a plain text diagram. The 2 failures in our run underline that cost even though the available log does not explain them. A team that only wants a sequence diagram in a README may get a faster, more portable result from Mermaid. Archify earns its overhead when interaction and verification are actually used.

Fresh releases and pushes indicate an active project

The repository has 16,851 stars and 21 open issues, substantial adoption signals for a developer tool. Activity is current: v2.15.0 was released on August 17, 2026, and the repository was pushed on August 26, 2026. The 2.16.0-dev.0 development badge fits that timeline. Stars do not prove support quality, and an issue count does not reveal response speed, but the release and push dates together point to ongoing maintenance rather than a popular abandoned repository.

Documentation is another strength. The English README links a Simplified Chinese version and includes agent-specific commands, bounded prompts, an 11-scenario Proof Lab, checked artifacts, export explanations, and a source-traced case study. That is enough to judge the product shape before installation. It does not erase our 708 of 710 result, so teams should still run the checks in their own supported environment and capture the actual failing test names.

It fits beside source control and design review

Archify belongs after repository analysis and before a map is accepted into documentation. An agent reads code or a description, Archify turns the findings into a checked artifact, and engineers review the topology beside a pull request or proposal. The HTML can live in internal docs, static hosting, or release assets, while PNG and SVG cover less interactive destinations. It does not replace tracing, metrics, service catalogs, or runtime discovery.

For teams already using 1 supported coding agent, the trial cost is low enough for a bounded experiment. Choose a system with known routes, request one primary path and explicit trust boundaries, then compare the result with source and operational knowledge. Adopt it if revision links, route tracing, or Before, Delta, and After views shorten real reviews. Otherwise, keep the simpler text diagrammer. Archify is strongest when the map becomes a reviewed engineering artifact, not a decorative screenshot.

Alternatives

ProjectWhat it isPick it when
Mermaid gh↗Text-based diagrams that render in many documentation systems and code hosts.Pick this instead when broad Markdown integration matters more than an interactive, agent-produced system map.
D2A declarative diagramming language with attractive layouts and multiple export paths.Pick this instead when you want to author diagram source directly and control layout through a dedicated language.
DiagramsPython code for drawing cloud and infrastructure architecture diagrams.Pick this instead when your team wants diagrams as Python code and primarily maps infrastructure resources.
Structurizr DSLA DSL for C4-style software architecture models and views.Pick this instead when a long-lived C4 model and formal architecture practice matter more than chat-native generation.

What people are saying

  1. [github-trending] tt-a1i/archify

Sources

  1. Archify GitHub repository
  2. Archify project page

More dev tools reviews

v2rayNG · SecLists · hashcat · eslint · fastfetch · composer · the whole board →