Cryyer
Automated product update emails, with per-product voice powered by LLM-drafted content. Supports multiple audiences per product, multiple LLM providers (Anthropic, OpenAI, Gemini), and multiple email providers (Resend, Gmail).
Website: cryyer.dev
GitHub: atriumn/cryyer
Issues: Report a bug
Cryyer supports two pipelines:
Release pipeline (manual approval before send):
- Push to
main → release-please opens a version-bump PR
draft-email.yml generates a draft file (drafts/vX.Y.Z.md) and commits it to the PR branch
- Merge the PR → tag pushed →
release.yml publishes to npm → GitHub Release created
send-email.yml fires on release publish but pauses for approval — review the draft, then approve to send
Weekly pipeline (manual review via GitHub issues):
weekly-draft.yml runs on cron (Monday 1pm UTC) — gathers GitHub activity, generates LLM drafts, creates GitHub issues for human review
send-update.yml fires when a draft issue is closed — sends emails to subscribers
Quickstart
For humans — interactive wizard
mkdir my-updates && cd my-updates
npx @atriumn/cryyer init
The interactive setup walks you through product name, GitHub repo, voice/tone, LLM provider, subscriber store, and API keys — then creates everything you need:
products/*.yaml — product configuration
.env — API keys and settings
subscribers.json — subscriber list (when using JSON store)
.gitignore — ignores .env and data files
For CI — flags + env vars
npx @atriumn/cryyer init --yes \
--product "My App" \
--repo owner/my-app \
--voice "Friendly and concise" \
--pipeline weekly
Non-interactive mode (--yes or CI=true) skips all prompts. Secrets are read from environment variables (ANTHROPIC_API_KEY, RESEND_API_KEY, etc.) instead of being written to .env. Defaults: anthropic LLM, json subscriber store, resend email, no workflows.
Available flags: --product, --repo, --voice, --llm, --subscriber-store, --email-provider, --from-email, --pipeline (weekly, release, or both).
For agents — write files directly
AI coding agents (Claude Code, Cursor, etc.) can skip init entirely and write the files themselves. All you need is a products/*.yaml file — see Product Configuration for the schema.
Then:
npx @atriumn/cryyer check
npx @atriumn/cryyer run --dry-run
When you're ready to run for real:
npx @atriumn/cryyer run
Or run the two stages separately:
npx @atriumn/cryyer draft
npx @atriumn/cryyer send
For release-triggered emails, use the file-based commands:
npx @atriumn/cryyer draft-file --product my-app --version 1.2.0
npx @atriumn/cryyer send-file drafts/v1.2.0.md --product my-app
You can also create products/*.yaml files manually — see Product Configuration for all fields, and .env.example for all environment variables.
Subscriber Stores
Set SUBSCRIBER_STORE to choose your backend. Default is supabase.
JSON File (simplest)
SUBSCRIBER_STORE=json
SUBSCRIBERS_JSON_PATH | ./subscribers.json | Path to subscriber data |
EMAIL_LOG_JSON_PATH | ./email-log.json | Path to email send log |
File format — array of objects with email, optional name, and productIds:
[
{ "email": "alice@example.com", "name": "Alice", "productIds": ["my-app", "other-app"] }
]
Supabase (default)
SUBSCRIBER_STORE=supabase
SUPABASE_URL | Your Supabase project URL (https://[project-id].supabase.co) |
SUPABASE_SERVICE_KEY | Supabase service role key |
Expects a beta_testers table with columns: email, name, product, unsubscribed_at.
Google Sheets
SUBSCRIBER_STORE=google-sheets
GOOGLE_SHEETS_SPREADSHEET_ID | The ID from your spreadsheet URL |
GOOGLE_SERVICE_ACCOUNT_EMAIL | Service account email |
GOOGLE_PRIVATE_KEY | Service account private key (PEM format) |
Cryyer looks for a sheet tab named after the product ID (e.g. my-app), falling back to the first sheet. Expected columns: email, name (optional), unsubscribed (optional, set to true to exclude).
Email send logging is a no-op with this backend (read-only).
Google Sheets setup walkthrough
1. Enable the Google Sheets API
- Go to console.cloud.google.com (create a project if you don't have one)
- Navigate to APIs & Services > Library
- Search for "Google Sheets API" and click Enable
2. Create a service account
- Go to APIs & Services > Credentials
- Click Create Credentials > Service account
- Name it (e.g.
cryyer-sheets-reader) and click Done
3. Generate a key
- Click the service account you just created
- Go to the Keys tab
- Click Add Key > Create new key > JSON
- Save the downloaded file
4. Set your environment variables
From the downloaded JSON file, grab client_email and private_key:
GOOGLE_SERVICE_ACCOUNT_EMAIL=cryyer-sheets-reader@your-project.iam.gserviceaccount.com
GOOGLE_PRIVATE_KEY="-----BEGIN PRIVATE KEY-----\nMIIE...\n-----END PRIVATE KEY-----\n"
5. Get the spreadsheet ID
From the spreadsheet URL:
https://docs.google.com/spreadsheets/d/SPREADSHEET_ID_HERE/edit
^^^^^^^^^^^^^^^^^^^^
GOOGLE_SHEETS_SPREADSHEET_ID=1BxiMVs0XRA5nFMdKvBdBZjgmUUqptlbs74OgVE2upms
6. Share the spreadsheet
Share the spreadsheet with your service account email (from step 4) as an Editor (required for add/remove subscriber support; Viewer is sufficient for read-only use).
7. Set up columns
Row 1 should have these headers:
Name the sheet tab after your product ID (e.g. my-app), or just use the default first sheet if you have one product.
LLM Providers
Set LLM_PROVIDER to choose your LLM backend. Default is anthropic.
LLM_PROVIDER | anthropic | anthropic, openai, or gemini |
LLM_MODEL | Per-provider default | Override the default model |
ANTHROPIC_API_KEY | — | Required when LLM_PROVIDER=anthropic |
OPENAI_API_KEY | — | Required when LLM_PROVIDER=openai |
GEMINI_API_KEY | — | Required when LLM_PROVIDER=gemini |
Default models: Anthropic claude-sonnet-4-5-20250514, OpenAI gpt-4o, Gemini gemini-1.5-flash.
Product Configuration
Products are defined as YAML files in products/. Each file represents one product.
id | Yes | Unique identifier, also used as GitHub issue label |
name | Yes | Display name |
voice | Yes* | LLM voice/tone instructions (injected into the draft prompt) |
repo | Yes | owner/repo for GitHub activity gathering |
emailSubjectTemplate | Yes* | Subject line template — use {{weekOf}} for the date or {{version}} for the release version |
audiences | No | List of audience-specific overrides (see docs site) |
tagline | No | Product tagline |
from_name | No | Override sender name for this product |
from_email | No | Override sender email for this product |
reply_to | No | Reply-to address |
* Required when audiences is not set. When using audiences, voice and emailSubjectTemplate are set per-audience instead.
GitHub Actions Workflows
weekly-draft.yml
Runs every Monday at 1pm UTC. Gathers GitHub activity and creates draft issues.
Secrets needed: GITHUB_TOKEN, CRYYER_REPO, and the API key for your chosen LLM_PROVIDER.
send-update.yml
Fires when an issue with the draft label is closed. Sends emails to subscribers.
Secrets needed: GITHUB_TOKEN, email provider credentials (RESEND_API_KEY or GMAIL_REFRESH_TOKEN), FROM_EMAIL, plus the secrets for your chosen SUBSCRIBER_STORE.
draft-email.yml
Runs when a release-please--* PR is opened or synced. Generates drafts/vX.Y.Z.md via LLM and commits it to the PR branch.
release.yml
Runs on v* tag push (after release-please PR is merged). Typechecks, tests, publishes to npm, creates GitHub Release.
send-email.yml
Runs when a GitHub Release is published. Reads the draft file and sends emails to subscribers. Pauses for manual approval via a production environment with required reviewers.
ci.yml
Runs on push/PR to main. Lints, typechecks, and runs tests.
Composite Actions
Reusable composite actions for consumer repos: atriumn/cryyer/.github/actions/draft-file@v0 and atriumn/cryyer/.github/actions/send-file@v0. Run cryyer init to scaffold wrapper workflows. See the docs site for full input references.
MCP Server
Cryyer includes an MCP server that lets you review, edit, and send drafts conversationally from any MCP client. It also supports subscriber management.
The MCP server uses stdio transport and is available as a separate binary: cryyer-mcp.
Standalone usage
npx @atriumn/cryyer-mcp
Or if installed locally:
pnpm run build
node dist/mcp.js
Claude Desktop
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
{
"mcpServers": {
"cryyer": {
"command": "npx",
"args": ["@atriumn/cryyer-mcp"],
"env": {
"GITHUB_TOKEN": "ghp_...",
"CRYYER_REPO": "owner/repo",
"CRYYER_ROOT": "/path/to/cryyer",
"RESEND_API_KEY": "re_...",
"FROM_EMAIL": "updates@example.com",
"SUBSCRIBER_STORE": "json",
"LLM_PROVIDER": "anthropic",
"ANTHROPIC_API_KEY": "sk-ant-..."
}
}
}
}
Other MCP clients (Cline, Continue, Windsurf, etc.)
Any MCP client that supports stdio transport can use Cryyer. The generic config is:
- Command:
npx @atriumn/cryyer-mcp (or node /path/to/cryyer/dist/mcp.js)
- Transport: stdio
- Environment variables: same as above
Only GITHUB_TOKEN and CRYYER_REPO are needed for read-only tools (list_drafts, get_draft). Other env vars are needed for sending, regenerating, and subscriber management.
Debugging
Test the server interactively with the MCP Inspector:
npx @modelcontextprotocol/inspector node dist/mcp.js
All log output goes to stderr (stdout is reserved for JSON-RPC). If something isn't working, check stderr for error messages.
Tools
list_drafts | List open draft issues |
get_draft | Get full draft content with subscriber count |
update_draft | Save revised subject + body |
send_draft | Send emails, close issue, post stats |
regenerate_draft | Re-gather activity + re-draft via LLM |
list_products | Show configured products |
list_subscribers | Show subscribers for a product |
add_subscriber | Add a subscriber to a product |
remove_subscriber | Unsubscribe someone from a product |
Prompt
Use the review_drafts prompt for the Monday morning review workflow — it walks through each pending draft and asks whether to send, edit, regenerate, or skip.
Development
pnpm install
pnpm run build
pnpm run init
pnpm run check
pnpm run typecheck
pnpm run lint
pnpm test
pnpm run test:watch
pnpm run dev
pnpm run mcp