pray.rs

A quiet prayer book for the web.

Write prayers and praises, keep them private, share them with a group, or make them public — read one page at a time, the way you'd turn through a book. No feed, no algorithm, no performance.

pray.rs

---

What it is

pray.rs is a small, intentional alternative to social media for prayer. It is built around a few ideas:

• A book, not a feed. Prayers are paged through with arrow keys or swipe gestures — one at a time. There is no infinite scroll and nothing is ranked.
• You choose who sees each prayer. Every prayer is private, shared with a group you belong to, or public.
• Prayer groups. Invite people by email, text, or a QR/link to pray together.
• Bring your own AI. A built-in MCP server lets you connect an AI assistant (Claude Code, claude.ai, Gemini CLI) to read and write your prayers on your behalf — authenticated as you, scoped to what you allow.

Tech stack

Concern	Choice
Frontend + SSR	Leptos (server-rendered + WASM hydration)
Web server	Axum
Database	SQLite via SQLx (runtime queries, no DATABASE_URL at compile time)
Auth	OAuth2 (Google + Facebook) with tower-sessions
AI integration	Model Context Protocol server with an OAuth 2.1 authorization server (DCR + PKCE) and static API keys
Email	Resend for group invitations
Hosting	DigitalOcean droplet behind Caddy, systemd-managed

Repository layout

crates/
  core          Shared types (Post, User, Group, Visibility, …)
  db            SQLx SQLite layer — migrations + repositories
  app           The Leptos full-stack app (SSR binary + WASM hydrate lib)
  project-mcp   A separate stdio MCP server for tracking feature requests + changelog
deploy/         Deploy script, systemd unit, droplet setup

Getting started

Prerequisites

• Rust (stable) with the wasm32-unknown-unknown target:

  rustup target add wasm32-unknown-unknown

• cargo-leptos:

  cargo install cargo-leptos

Configuration

Copy the example env file and fill it in:

cp .env.example .env

Variable	Purpose
DATABASE_URL	SQLite path, e.g. sqlite://thanksgivings.db
BASE_URL	Public base URL (used in invite links)
GOOGLE_CLIENT_ID / GOOGLE_CLIENT_SECRET	Google OAuth
FACEBOOK_CLIENT_ID / FACEBOOK_CLIENT_SECRET	Facebook OAuth
RESEND_API_KEY / RESEND_FROM	Outbound invitation email
DEV_AUTH_BYPASS	Set to 1 to enable a local dev login that skips real OAuth

Run it

cargo leptos watch

Then open http://127.0.0.1:3000. With DEV_AUTH_BYPASS=1, use the dev login to sign in without configuring OAuth.

Check your work

cargo check alone won't catch everything — verify both build targets:

cargo check -p app --features ssr
cargo check -p app --features hydrate --target wasm32-unknown-unknown

See CLAUDE.md for project conventions and gotchas (notably: prefer extracting components over raising recursion_limit, and always bind SQL values rather than interpolating them).

Connect an AI assistant

In the app, go to Settings → Connect your AI. The recommended path signs in with your account over OAuth — no API key to manage:

claude mcp add --transport http prayers https://pray.rs/mcp

Then run /mcp in Claude Code, select prayers, and Authenticate (or Reauthenticate to move from public prayers to your own account). A static API key is also available as an alternative.

Deploying

bash deploy/deploy.sh

This cross-compiles the Linux server binary with cargo zigbuild, builds the optimized WASM/CSS with cargo leptos build --release, uploads to the droplet, and restarts the service. Requires zig and cargo-zigbuild.

License

© Andrew Gauger. Personal project — all rights reserved unless noted otherwise.