Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
Write whole barbell programmes, not just workouts: weeks, progressions, ramps, and plate maths.
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.
You need
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Works in Claude on the web, desktop and mobile:
https://tectonicplates.app/mcpRun this in your terminal
claude mcp add --transport http tectonic-plates https://tectonicplates.app/mcpPaste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json
{
"mcpServers": {
"tectonic-plates": {
"url": "https://tectonicplates.app/mcp"
}
}
}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": {
"tectonic-plates": {
"type": "http",
"url": "https://tectonicplates.app/mcp"
}
}
}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.
ChatGPT connects only to servers on the internet:
https://tectonicplates.app/mcpMenu names can differ slightly by plan and over time.
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.
We have not read this server’s tool list yet; we check endpoints in rotation.
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 7, 2026
The publisher (tectonicplates.app) is not the repo owner (stephaniewilkinson); someone else may have listed it
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
Not checked yet; we check endpoints in rotation
Not checked yet; we check endpoints in rotation
Not checked yet; we read the tool list when we check the endpoint
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 0 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.
The registry does not list a required key, though some servers ask you to sign in on first use.
It runs in the maker’s cloud. Connect over the internet with nothing to install, but what you send passes through the maker’s servers.
The GitHub repo last changed –, and the latest version (0.1.0) reached the registry 2 months ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. It connects by URL, so apps that accept remote MCP servers work too, such as ChatGPT in developer mode.
In the same category: Spec Driven Development, Remote Desktop Commander, AvaHR
Servers in the same category from other makers, sorted by stars.
Reading a card:★ 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.
Read hiring data and create inactive assessment and questionnaire drafts for human review.
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.
Tectonic is a barbell strength-training tracker. An account logs workouts; a workout holds sets of an exercise at a weight and a rep count, in the order they are lifted. All weights are integer pounds.
What makes it more than a list is that a session is written before it is lifted. A program expands into ordinary set rows — warmup ramp included, every weight rounded to something a bar can actually be loaded to — so lifting is tapping through a list that already knows what comes next, and each row keeps the prescription beside what was actually done. An MCP endpoint exposes the same data to an LLM as a connector, which is the second half of this file.
The stack is Roda and Sequel on Postgres, with Rodauth for accounts and rodauth-oauth for the OAuth 2.1 authorization server the MCP endpoint authenticates against. Views are ERB with Tailwind from a CDN, and htmx for the parts of the session screen that have to change without a page load. It deploys to Render.
/workouts/:id/session) — the gym-floor screen, as distinct from /workouts/:id, which is the record of one. A progress bar across the top, sets grouped by lift with warmups dimmed, the per-side plate breakdown under each weight ("per side 1×25 1×10"), and a Done button per set that toggles, so a mis-tap is undone by tapping again. A set lifted differently is entered inline and the row turns amber rather than lime, because done-as-written and done-differently have to read differently. The whole session gets an RPE rating, with a bar-speed guide for choosing one./mcp.You need Ruby at the version in .ruby-version (the Gemfile reads the same file, so the two cannot drift), a local Postgres, and — only for the browser specs — Firefox with geckodriver.
bundle install
npm install # only for the stylesheet; see below
dotenv loads .env when the app is required, and .env-example is a file to copy: cp .env-example .env gives a working development setup, pointed at the tectonic_development database rake db:create makes below. The keys that are secrets in a deployment are commented out there rather than left blank, each with the command that generates it, because a blank is worse than an absence here — dotenv sets the name to the empty string, every fallback in this project tests for nil, and an empty DATABASE_URL reaches Sequel.connect and raises rather than falling back to a default.
| Variable | Required | Notes | ||
|---|---|---|---|---|
DATABASE_URL | yes | app.rb connects at require time, so nothing loads without it. .env.rb, which only the Rakefile reads, defaults it from RACK_ENV to postgres:///tectonic_development or postgres:///tectonic_test, but it defaults with ` | =, so a value in .env wins and that default never fires. A test run is the exception: spec_helper names its own database whatever the environment already held, so a .env` cannot reach the suite. | |
SESSION_SECRET | yes | At least 64 bytes. Roda's sessions plugin refuses a shorter one, and app.rb builds the app at require time, so a short or missing secret raises before a single route is reached. .env-example carries one that is long enough and says in its own text that it is for development; generate a real one with ruby -rsecurerandom -e 'puts SecureRandom.hex(64)' for anything deployed. | ||
RACK_ENV | no | development unless set, and development is the only value that turns Sequel's query log on; production and staging initialise Sentry and require real OAuth keys. | ||
DB_LOG | no | Attaches Sequel's query log to stdout whatever RACK_ENV says — set it to anything non-empty to trace queries against a deployment, and unset it again afterwards. Sequel logs statements with their bound values, so those lines carry email addresses and everybody's weights and reps, which is why nothing but development logs by default. | ||
SENTRY_DSN | no | The Sentry project DSN, read only when RACK_ENV is production or staging. Without it — unset or empty — the app boots and serves with error reporting switched off and says so once on stderr: losing error reporting is not a reason to refuse to start, which is why this behaves unlike OAUTH_JWT_PRIVATE_KEY. | ||
RACK_TIMEOUT_SERVICE_TIMEOUT | no | Seconds a request may hold a thread before rack-timeout raises inside it, 20 unless set, 0 to switch it off — which is what you want locally the moment you stop in a debugger. Read only by config.ru, so the suite never sees it. It covers both legs of the URLMap, MCP included; the MCP endpoint's long-lived subscriptions/listen streams are unaffected either way, because rack-timeout times app.call and a streaming body is written after that has returned. | ||
RAILS_MAX_THREADS | no | 5 unless set, and read twice: config/puma.rb gives Puma that many threads and lib/tectonic/db.rb gives Sequel that many connections. One variable because the pool used to be Sequel's default of four against Puma's five threads, and a fifth request thread waiting on a connection is a Sequel::PoolTimeout on a request that had nothing wrong with it. The name is the host's convention rather than a claim about Rails. | ||
WEB_CONCURRENCY | no | Puma workers, 0 unless set — one process, which is what rackup was already running, and as much as half a CPU has to offer. Above zero Puma forks, so the app is preloaded and each worker disconnects Sequel on boot rather than sharing the parent's connections. Every worker multiplies the connection count by RAILS_MAX_THREADS. It is also what switches worker_timeout on: Puma enforces that in cluster mode alone, so at the default of zero nothing reaps a request still running after the response was handed back, and RACK_TIMEOUT_SERVICE_TIMEOUT is the only timeout in play. |
The MCP and OAuth variables are all optional in development and are documented in the table further down.
bundle exec rake db:create # createuser tectonic, then the development and test databases
bundle exec rake db:migrate # applies migrate/001_schema.rb against DATABASE_URL bundle exec rake library:exercises # loads the 54 built-in movements, idempotent on name
db:create shells out to createuser -U postgres and createdb -U postgres -O tectonic, so it assumes a superuser role named postgres and gives the databases to a tectonic role. If your Postgres has neither, createdb tectonic_development by hand and go straight to db:migrate, which only needs DATABASE_URL to point somewhere it can connect.
The schema is one squashed baseline. db:migrate stamps a database that already carries the tables rather than trying to rebuild or roll it back, so an existing database from before the squash adopts the baseline instead of breaking on it.
bundle exec rackup config.ru # http://localhost:9292
config.ru mounts two apps side by side under Rack::URLMap: the MCP endpoint at /mcp, which has its own bearer-token auth and never touches Roda's sessions or CSRF, and the Roda app at everything else. / redirects to /welcome until you are logged in, so start by signing up at /create-account.
npm run build:css # assets/css/app.css -> assets/css/styles.css
assets/css/styles.css is compiled from assets/css/app.css and committed, so nothing but a stylesheet edit needs Node at all — bundle install and a database are enough to run the app and the suite.
Rebuild it after adding a Tailwind class that was not already in use. CI rebuilds and fails on a difference, so a stale file is caught before it ships rather than after: a purged class breaks nothing anywhere except on the page it was meant to style, in production, silently.
tailwind.config.js scans app.rb and lib/**/*.rb as well as the templates, because button_style, rpe_style, row_style and the calendar's STYLES table build class names in Ruby and appear in no template.
It used to be — the Play CDN, which compiles in the browser. That meant no CSS at all with JavaScript off, and a third party on every page including the consent screen.
bundle exec rake test # the whole suite
bundle exec ruby -Ispec spec/set_scheme_spec.rb # one file
Neither variable has to be passed in. spec_helper sets RACK_ENV and then names the database itself — TEST_DATABASE_URL when a run wants one of its own, and postgres:///tectonic_test otherwise — whatever the shell, a .env or the Rakefile had already chosen. That matters more than it looks: spec_helper empties every table after every test, so a suite pointed at your development database would not seed it, it would clear it. The built-in exercise library is the one thing kept, because a deployed database holds it before anybody signs up. rake 'db:reset[name]' still rebuilds a database whose schema has gone wrong, but a run no longer leaves rows behind for it to clear away.
Two prerequisites, neither obvious from the failure you get without them:
app.rb connects at require time, so even the pure unit specs need a live database as soon as spec_helper loads the app.spec/exercises_spec.rb, route_ownership_spec.rb, session_spec.rb, system_spec.rb and workouts_spec.rb drive a browser through Capybara and selenium, on a Puma server bound to port 9292. MOZ_HEADLESS=1 keeps windows from opening; CI sets it for the same reason.The linters, both of which CI runs — these exact commands, not equivalents of them:
bundle exec rubocop
bundle exec rake lint:views
lint:views is a rake task rather than erb_lint with a glob because the glob had to be expanded by a shell, which reached only one directory deep and was refused outright in several sandboxed environments. The Rakefile builds the file list in Ruby, so the list is readable, recursive, and the same one in CI as on your machine.
A program is a plan; workouts and sets are what it produces. Four tables hold the plan — programs, program_weeks, program_days, program_lifts — and generating a week turns them into ordinary workouts and sets rows with is_completed false. A planned session is not a separate kind of record: lifting a set as written only flips is_completed, while lifting it differently changes weight or reps and leaves planned_weight and planned_reps behind as the record of what was asked for.
start_date, and two preferences the maths below reads: preferred_reps (the rep count main work should be expressed at) and is_ascending (whether working sets climb to the top weight or sit flat).is_deload. Dates are derived rather than stored: a week opens seven days after the one before it, counted from the block's start.sets, reps, is_barbell (whether it warms up off a bar andOnly the beginning is shown; read the rest in the repo