Archify teaches your coding agent to draw. Describe a system in a sentence, or point the agent at a repository, and it writes the diagram as structured data, runs it through validators and renders one standalone HTML file. You can open that file in a browser, trace a path through it, focus on one node, switch themes, or export it as an image or video. Whoever you send it to needs nothing installed.
It covers ten diagram types: architecture, workflow, sequence, data flow, lifecycle, ERD, class, tree, timeline and waterfall. It can also answer a question as a one-page explainer, with the conclusion first and tables, steps and diagrams underneath. It is not only for code; the project's gallery includes a four-day Shanghai walking itinerary a community member made with it. You refine the result in chat with requests like add authentication or highlight the cache-miss path.
Archify works with Claude Code, Codex, Cursor, OpenCode and GitHub Copilot, among others. It is Apache-2.0 licensed, with about 94,000 stars and 92 contributors since April 2026, and version 3.2.0 is the latest release. The README also says it has passed 130,000 installs on skills.sh.
- Repository: github.com/tt-a1i/archify
- Licence: Apache-2.0 (Apache License 2.0)
- Language: JavaScript. Stars: 94.3K. Forks: 6,420. Last push: Nov 7, 2026.
- Scan: safe, Nov 7, 2026, commit a406594
Who it is for
Developers documenting a system or explaining a pull request, tech leads preparing a design review, new team members mapping a codebase, and anyone who already uses a coding agent and wants a diagram that people can click through.
Getting started
1. Paste this into your agent. It reads the project's short install guide and should confirm the scope with you first
Install Archify for me by reading https://raw.githubusercontent.com/tt-a1i/archify/main/llms.txt2. Or install it for Claude Code yourself (Node 18 or later)
npx -y skills@latest add https://archify.si --skill archify --agent claude-code --global --copy --yes3. Then ask your agent for a diagram
Use Archify to diagram a web request: Browser calls the API, the API checks Redis, and a cache miss queries PostgreSQL and fills the cache.4. Optional: switch off the daily update check
export ARCHIFY_UPDATE_CHECK_DISABLED=1Diagrams are only as true as what the agent writes. Archify's validators check structure and layout, not whether the system really works that way, so review anything drawn from a real codebase before you rely on it.
Safety scan
We cloned tt-a1i/archify at commit a406594 on Nov 7, 2026 and ran the checks described on the GitHub Tools page: credential patterns, decode-and-execute code, install-time scripts, committed binaries, risky CI workflows, every host the code talks to, known vulnerabilities in pinned dependencies, and project hygiene. A person read every hit. This is what we found.
- One secret hit across 1,189 files and about 1.1 million lines, mostly generated HTML, and it is not a secret: test/explainer-diff.test.mjs line 85 holds only the text BEGIN PRIVATE KEY with no key, to prove that Archify's own secret scanner rejects it. No pattern hits. The three IP addresses are reserved documentation addresses in a test.
- It sends very little. There is no telemetry. During delivery the skill makes a plain GET for a fixed version file on tt-a1i.github.io, about once a day, and only shows a reminder; it never installs anything. The request carries no version, project data or prompts, though the server sees your IP and the time, as the README says. Set ARCHIFY_UPDATE_CHECK_DISABLED=1 to stop it. The diagram template has no network calls or external scripts, and repository evidence comes from your local git.
- The extra pass for skill repos was clean. No hook files or agent settings changes; the only mention is a design note saying a session hook is not implemented. No exfiltration hosts and no pipe-to-shell. The one injection phrase is a string in a test. The only hidden character is an emoji joiner in a test. The only credential paths are in a deny-list that keeps .env, .ssh and .aws files out of source excerpts. The .agents folder holds the maintainers' own skills, which the documented install does not add.
- The README carries sponsors: Kimi Work, Supercode, OpenLux, PackyCode, AICodeMirror and Infistar, all with referral or tracking links, and we did not review them. Several advertise access to Claude, Codex or Gemini through their own API at a small share of official prices; anything your agent sends through a service like that passes through its operator. None of these names appear in the skill itself, so they do not reach your agent or your diagrams. The project says those services are run by their providers and do not represent Archify.
- Advisories: website/package-lock.json has 22, with eight high: sharp, source-map-js, four in undici, one in Vite and one in Zod. That lockfile builds the project's static website and is not part of the skill. The skill's own lockfile (five, moderate and low) and the root one (four, moderate and low) are all fast-uri, used by development tools. The skill's package.json lists only development dependencies, so installing it brings in no npm packages.
- archify.zip at the top of the repo is the upload bundle for Claude.ai. We opened it: 202 of its 203 files are byte-identical to the archify folder, and the other is package.json with its development scripts and dependencies removed.
- Six workflows with 62 action uses, and all 11 third-party uses are pinned to commits. One uses pull_request_target, to make a thank-you card when a pull request is merged, and it checks out the default branch, never the pull request's code. A licence, contributing guide and security policy are present; no Dependabot.
What the scanner counted
| Check | Result |
|---|---|
| Secrets | 1 candidate found and read; see the notes above. |
| Suspicious code | None found. |
| Install-time code | None: nothing runs at install beyond the package manager itself. |
| Committed binaries | None. |
| CI workflows | 6 workflows. 1 uses pull_request_target, none check out the pull request head. 0 of 11 third-party actions pinned to a tag rather than a commit. |
| Network hosts | 35 distinct hosts referenced from source; most often github.com, archify.si, git.internal, tt-a1i.github.io. 3 URLs to a bare IP address, listed under the raw findings. |
| Known vulnerabilities | 27 advisories across 459 pinned packages: 0 critical, 8 high, 14 moderate, 5 low. archify/package-lock.json: 6 packages, 5 advisories; package-lock.json: 17 packages, 4 advisories; website/package-lock.json: 450 packages, 22 advisories. |
| Project hygiene | Has security policy, licence file, contributing guide. Missing automated dependency updates, CodeQL. |
| OpenSSF Scorecard | Not scored: the project is not in Scorecard's weekly index. |
The raw findings
Every hit the scanner wrote out, with a link to the exact line at the scanned commit. Secrets candidates are redacted.
Secret candidates (1, redacted)
| Where | Rule | Match |
|---|---|---|
| test/explainer-diff.test.mjs:85 | private-key | -----B…--- (27 chars) |
URLs to bare IP addresses (3)
| Where | Rule | Match |
|---|---|---|
| test/brand-marks.test.mjs:537 | ip-literal-url | 'http://192.0.2.1/', |
| test/brand-marks.test.mjs:538 | ip-literal-url | 'http://198.51.100.1/', |
| test/brand-marks.test.mjs:539 | ip-literal-url | 'http://203.0.113.1/', |
Worst known vulnerabilities (24 of 27)
| Advisory | Severity | Package | Summary |
|---|---|---|---|
| GHSA-wq5f-xc86-pv6w | high | sharp@0.35.4 | sharp : Vulnerability in librsvg dependency CVE-2026-96889 |
| GHSA-68fv-2mgg-jv7q | high | source-map-js@1.2.1 | source-map-js allows event-loop denial of service through indexed source-map section offsets |
| GHSA-2m5v-pch6-74hx | high | undici@8.10.2 | undici vulnerable to TLS certificate validation bypass via dropped tls options in Agent |
| GHSA-4w7w-5xrc-9h26 | high | undici@8.10.2 | undici vulnerable to cache poisoning and cross-user disclosure via cross-origin redirects |
| GHSA-qrxm-fqhp-jwvm | high | undici@8.10.2 | undici vulnerable to Denial of Service via TLS 1.2 certificate-changing renegotiation |
| GHSA-xjj6-f33q-929c | high | undici@8.10.2 | undici vulnerable to Denial of Service via unhandled error on excessive content-encodings |
| GHSA-587h-7wmj-h5pm | high | vite@8.3.0 | launch-editor: `fs.existsSync()` existence check can be bypassed on Windows |
| GHSA-mrhx-c7xc-m3w5 | high | zod@4.6.5 | Zod can exhaust memory while validating oversized constrained collections |
| GHSA-hrr3-gc8f-f4qj | moderate | fast-uri@3.1.7 | fast-uri vulnerable to inconsistent host case normalization via percent-encoded octets |
| GHSA-w74x-v6f9-j3xg | moderate | fast-uri@3.1.7 | fast-uri vulnerable to URI scheme spoofing via percent-encoded scheme decoding |
| GHSA-wjr9-j795-jxcr | moderate | fast-uri@3.1.7 | fast-uri vulnerable to host confusion via unchecked WHATWG host canonicalization |
| GHSA-w74x-v6f9-j3xg | moderate | fast-uri@3.1.8 | fast-uri vulnerable to URI scheme spoofing via percent-encoded scheme decoding |
| GHSA-wjr9-j795-jxcr | moderate | fast-uri@3.1.8 | fast-uri vulnerable to host confusion via unchecked WHATWG host canonicalization |
| GHSA-r4xh-jqrq-34v2 | moderate | smol-toml@1.8.0 | smol-toml: Quadratic-time parse() from parseKey rescanning to end of document on each key line |
| GHSA-4qv8-2p9j-2685 | moderate | undici@8.10.2 | undici vulnerable to Denial of Service via unbounded memory in the EventSource parser |
| GHSA-55wc-3pg8-jfw6 | moderate | undici@8.10.2 | undici vulnerable to Denial of Service via unhandled exception in the decompress interceptor |
| GHSA-9fm3-97px-rjf6 | moderate | undici@8.10.2 | undici WebSocket client vulnerable to Denial of Service via unbounded automatic PONG buffering |
| GHSA-f72v-6w45-q472 | moderate | undici@8.10.2 | undici vulnerable to Subresource Integrity bypass via non-lowercase algorithm tokens |
| GHSA-698v-4r3x-3vcv | moderate | vite@8.3.0 | launch-editor: escapeCmdArgs() omits % escaping on Windows |
| GHSA-r825-w8mm-5pxw | moderate | vite@8.3.0 | launch-editor: Editor CLI flag misinterpretation causing arbitrary command execution for some editors |
| GHSA-rq7h-c2jc-7f22 | moderate | vite@8.3.0 | Vite: Files outside `server.fs.allow` could be served when project paths match system paths on non-Windows machine |
| GHSA-vfpm-58rq-9qcg | moderate | vite@8.3.0 | Vite: First 4 bytes of files outside of `server.fs` setting is exposed by `?vite-wasm-instance` |
| GHSA-333x-w4r6-6crp | low | fast-uri@3.1.7 | fast-uri vulnerable to scheme injection via an unescaped first-segment colon in suffix references |
| GHSA-xhq7-xmqj-769h | low | fast-uri@3.1.7 | fast-uri vulnerable to improper path equivalence in relative reference normalization |
Workflows worth a look
- .github/workflows/contributor-cards.yml: pull_request_target
By the numbers
| Stars | 94.3K |
|---|---|
| Forks | 6,420 |
| Contributors | 92 |
| Commits | 1,759 |
| Open issues | 143 |
| Open pull requests | 179 |
| Releases | 23 |
| Latest release | v3.2.0 |
| Licence | Apache-2.0 |
| Main language | JavaScript |
| Project age | 6 months |
| Last push | Nov 7, 2026 |
| Tracked files | 1,189 |
| Lines of code | 1.1M |
| Checkout size | 104 MB |
Lines by language: HTML 850.5K, JavaScript 162K, JSON 55.1K, Markdown 29.1K, CSS 7,288, YAML 1,586.
Questions
Is Archify free?
Yes. It is free and open source under Apache-2.0, with no account. Your only cost is your agent's usage. The paid services in its README are separate companies, and you do not need any of them.
Does Archify need a code repository?
No. You can start from a plain description, a question or a plan. If you do ask it to read a repository, it can pin source references to a commit so you can check where each part came from.
Can I share a diagram with someone who does not have Archify?
Yes. The output is one HTML file, so they can open it in a browser and keep the interactions. Only external links need a network connection. You can also export images, video and share cards.
Do I have to use the sponsors' API services?
No. Archify runs inside whichever agent you already use and makes no calls to them. They are sponsors of the project, not part of the tool. If you consider one, check its terms and data handling yourself, as we did not review them.
Is it a Mermaid replacement?
Not exactly. It does not parse Mermaid automatically, and it is not a drag-and-drop editor or a hosted sharing service. It is for an agent generating a polished interactive page that you then adjust by asking in chat.
This post is part of GitHub Tools, where every repository is cloned and scanned before it is written up. The scan is a snapshot of one commit on one day; the repository has moved on since, so check it before you install.