Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
Self-hosted catalogue of your skills, agents and MCP servers, searchable by meaning over MCP
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
SECRET_KEY, GOOGLE_API_KEYValues marked secret are keys to your account. Never share them or paste them into a chat.
You need
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Open Claude Desktop → Settings → Developer → Edit Config, and paste this into claude_desktop_config.json
{
"mcpServers": {
"vivatlas": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"SECRET_KEY",
"-e",
"GOOGLE_API_KEY",
"ghcr.io/bobpanil/vivatlas:1.0.13"
],
"env": {
"SECRET_KEY": "your-value",
"GOOGLE_API_KEY": "your-value"
}
}
}
}Run this in your terminal
claude mcp add vivatlas --env SECRET_KEY=your-value --env GOOGLE_API_KEY=your-value -- docker run -i --rm -e SECRET_KEY -e GOOGLE_API_KEY ghcr.io/bobpanil/vivatlas:1.0.13Fill in the empty values before running it.
Paste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json
{
"mcpServers": {
"vivatlas": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"SECRET_KEY",
"-e",
"GOOGLE_API_KEY",
"ghcr.io/bobpanil/vivatlas:1.0.13"
],
"env": {
"SECRET_KEY": "your-value",
"GOOGLE_API_KEY": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
Paste into .vscode/mcp.json in your project
{
"servers": {
"vivatlas": {
"type": "stdio",
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"-e",
"SECRET_KEY",
"-e",
"GOOGLE_API_KEY",
"ghcr.io/bobpanil/vivatlas:1.0.13"
],
"env": {
"SECRET_KEY": "your-value",
"GOOGLE_API_KEY": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.
Your choice stays on this device in a cookie; no account needed.
These settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.
Runs on your computer; the tool list appears in your AI app after you install it.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 8, 2026
The registry name was verified with the GitHub account bobpanil, which owns the repo
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
This one runs on your computer; there is no internet endpoint
Runs on your computer and uses local settings instead
Runs on your computer; we can only read tool lists from internet endpoints
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
You need to fill in: SECRET_KEY, GOOGLE_API_KEY
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
The GitHub repo last changed –, and the latest version (1.0.13) reached the registry 23 days ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Docker on your computer.
In the same category: Spec Driven Development, Remote Desktop Commander, aamio rendezvous
Servers in the same category from other makers, sorted by stars.
Reading a card:↓ Downloads per weeki★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Meet an agent you have not met, exchange messages that expire, and prove it happened. No account.
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
You collect tools. Skills for Claude, agents, MCP servers, a scraper you starred at midnight, a design kit somebody linked on Reddit. They end up scattered across GitHub stars, a self-hosted Gitea, browser bookmarks and a notes app — and three months later you remember *there was a thing for that* and can't find it.
VIVATLAS is the shelf. Point it at your repositories and the links you clip, and it turns each one into a card: a name, a description written for a person rather than a compiler, a picture, tags. Then it puts them all in one place you can search by *meaning*, not just by name. Ask it for "something that audits a site for accessibility" and it hands you three candidates and tells you why.
It runs on your own machine — a NAS, a spare box, a laptop — it's built for more than one person, and it only ever *reads*. Public repositories in, nothing written back to Git, ever.
Try the live demo
Try it before you install it. demo1.bobpanil.com is a public VIVATLAS. Press Try it on the sign-in page for a private account of your own, then browse the catalogue, search it, and add links or notes. The account and everything you add are deleted after 24 hours. The demo runs without AI, so the cards you add are plain and search goes by words, not by meaning. It always runs the newest release.
Nobody browses a wall of grey rectangles. VIVATLAS finds each project's own picture — the banner at the top of its README, a logo in the repo, or the card GitHub draws for it — and throws out the things that only *look* like pictures: build badges, sponsor logos, "buy me a coffee" buttons, the owner's avatar. When a project has several real candidates, the same AI that writes the description looks at them side by side and picks the one that actually shows the thing.
And when a project has nothing at all, VIVATLAS draws a cover itself — the name set in the house typeface on a colour and a pattern that come from the name, so every cover is different and they all belong together. No model, no cost, never grey.
Open one and you get the same project three ways: a one-liner, a paragraph for a person, and a technical note for when you're deciding whether to use it. Tags with their confidence, what it's for, where it came from and when it last changed — and the buttons to rescan it, file it, or copy the link. The descriptions are written by whichever AI you've configured, from the repository's own README.
Two indexes work together: full-text for words, vectors for meaning. Type in Russian and find an English tool. Describe a task instead of naming a tool, and the recommender returns three options with a sentence each on why — or an honest "nothing here fits", which is rarer in software than it should be.
Installing the Android app: from F-Droid, which keeps it updated. Or grab vivatlas-android- .apk from the latest release, or add this repository to Obtainium and it updates itself. Releases are signed with one key for the life of the app, so each installs over the last.
Each person has their own sign-in, their own private cards and personal folders. The shared catalogue is common to all and curated by whoever runs the place. A card can move between the two — it's the owner's call — and you can keep favourites, watch a feed of what changed, and see which cards have gone stale upstream.
VIVATLAS is an MCP server. Connect it to ChatGPT, Claude, or anything else that speaks MCP, and "which of my tools does X?" gets answered from *your* catalogue. There's a plain REST API too. → docs/MCP.md
Three languages (English, Russian, Hebrew — right-to-left done properly). Light, dark, OLED and follow-the-system themes. Two-step sign-in with backup codes. Invitations by link or email, or open registration if you prefer. Gitea and GitHub as sources, rescanned daily. An admin panel for people, folders, email and AI keys — all changeable without a restart. And upstream / update, which compare a card with where it came from and bring in a new version only where you haven't touched the file.
The quickest way is the container:
docker run -d --name vivatlas -p 8710:8710 -v vivatlas_data:/data ghcr.io/bobpanil/vivatlas:latest
Open http://localhost:8710, and the first person through /setup becomes the owner. Then Admin → Sources: give it a GitHub account or a Gitea address, press *Scan now*, and watch the cards arrive.
Running from source instead:
python -m venv .venv
.venv/bin/python -m pip install -e ".[dev]" # Windows: .venv\Scripts\python.exe cp .env.example .env # set SECRET_KEY; everything else is optional
.venv/bin/python -m vivatlas.cli init-db # create or update the database .venv/bin/python -m vivatlas.cli scan # fetch repositories .venv/bin/python -m vivatlas.cli serve --host 0.0.0.0 --port 8710
--host 0.0.0.0 lets a phone on the same Wi-Fi reach it. After updating the code, run init-db again — serve doesn't migrate on its own (the container does, on every start).
For a proper home, there's a walkthrough for TrueNAS SCALE in docs/DEPLOY-TRUENAS.md — it also covers the two settings that matter once VIVATLAS sits behind a proxy or a tunnel.
Everything optional lives in .env (or the container's environment); .env.example annotates the lot. The ones people ask about:
| Setting | What it's for |
|---|---|
GOOGLE_API_KEY, OLLAMA_URL | The AI that writes descriptions and picks pictures. Either works; both is fine. Also settable in the admin panel. |
TRUSTED_PROXIES | If Cloudflare, Traefik or nginx terminates TLS in front of VIVATLAS. Without it, cookies go out without Secure. |
IMAGE_MODEL | Have pictures *drawn* for cards that offer none — a Google image model (needs billing) or pollinations:flux (free, keyless, soft). Empty means the designed covers, which is the default and looks better than you'd think. |
VIVATLAS is a viewer, not an owner. It reads public Git repositories and helps you find and view the tools inside them. It does not own — and claims no rights in — any of the repositories, code, names or trademarks it catalogues or links to. Those belong to their authors and are governed by their own licences.
It writes nothing to Git and does not scan private repositories. That's a rule, not a setting: there's no toggle for it anywhere.
Licence: the server — everything under src/ — is Business Source License 1.1: free to use, deploy and modify, including commercially; you may not sell or resell VIVATLAS itself as a product or a hosted service. It converts to Apache-2.0 on 2030-07-21. The Android app (android/) is Apache-2.0, so it can be built and carried by F-Droid. The browser extension (extension/) is Apache-2.0 too.
.venv/bin/python -m pytest
.venv/bin/python -m ruff check src tests
Under the hood — where things live, for when you want to change something
src/vivatlas/
config.py settings from .env runtime_settings.py overrides from the database on top of .env (edited from the admin panel) db.py connection to SQLite models.py tables migrate.py init-db: add missing columns/indexes, rebuild search security.py passwords, secret encryption, secret key twofactor.py two-step sign-in (TOTP + backup codes) auth.py, auth_web.py sign-in, registration, invitations, password reset qrlogin.py sign a phone in by a one-time QR shown on a signed-in browser admin_web.py admin panel (people, access, email, integrations) settings_web.py personal settings, avatars, sources, folders web.py catalogue, cards, adding api.py app assembly, REST, background loops, static filters.py visibility: own + shared categories.py folder permissions (shared/personal) i18n.py, translations*.py translations (en/ru/he), RTL mailer.py emails (password reset, invitations) avatars.py uploaded photo → square WebP previews.py the picture on a card: find it, fetch it, fit it — or draw a cover ai/ the models: google.py (text, vision, images), ollama.py, pollinations.py usericons.py default avatar set (static/usericons) scanner.py, indexer.py scanning + the private-repo rule, index mcp_server.py MCP server for AI assistants cli.py terminal commands (init-db, scan, serve, previews, embed, upstream…) providers/ base.py common interface to a Git host (the "socket") gitea.py Gitea github.py GitHub (an account's public repositories) ext_api.py JSON API for the browser extension and the Andr
Only the beginning is shown; read the rest in the repo