Zero-setup WhatsApp notifications + human-in-the-loop for AI agents — text 'join', send in 60s.
WhatsApp notifications for AI agents and scripts with zero setup — no Meta
account, no templates, no dashboard: text "join", get an API key,
curl → WhatsApp in 60s. Two-way human-in-the-loop: your agent asks, you tap
a button to answer. From your terminal or an MCP client.
This is the open-source (MIT) pingwa client: a CLI, a tiny HTTP client, and an MCP server. It talks to a pingwa backend (the hosted service at https://pingwa.dev, or your own). The backend is separate and private; this client is not.
No install needed with uv:
export PINGWA_KEY=pw_your_key # get one: send "join" on WhatsApp to the service number
uvx pingwa send "build finished ✅" # notify your own phone
echo "deploy done" | uvx pingwa send - # from stdin
uvx pingwa ask "Deploy to prod?" --button yes --button no # wait for a reply from your phone
uvx pingwa replies # pull messages your phone sent back (out-of-band steering)
uvx pingwa me # plan / quota / reply-window
uvx pingwa keys # list active API keys (up to 10 named keys per number)
uvx pingwa keys revoke ci # revoke one by name (or reply 'revoke ci' on WhatsApp)
uvx pingwa upgrade # get a Stripe link to go Pro (no password needed)
uvx pingwa send "x" --json # raw JSON response
AI agents: don't ask your human to copy-paste a key — drive the registration
yourself (POST /v1/registrations; full recipe in
llms.txt). The human only taps one WhatsApp link;
the key is claimed by you and never travels through the chat. Works for existing
accounts too — it mints an additional named key.
Exit codes: 0 ok · 2 missing key · 3 quota exceeded · 1 other.
Override the backend with PINGWA_BASE_URL or --base-url.
Prefer PUSH over polling replies? Register HTTPS endpoints and pingwa POSTs
every inbound WhatsApp message to them:
uvx pingwa webhooks # list your webhooks
uvx pingwa webhooks add https://your-host/pingwa-hook # returns a signing SECRET, shown ONCE
uvx pingwa webhooks rm https://your-host/pingwa-hook # by id, or by exact url
add prints a whsec_... secret once — store it now; it never appears
again. Each delivery carries an X-Pingwa-Signature: sha256=<hex> header (an
HMAC-SHA256 of the raw body under that secret) — verify it and reject anything
that fails. Full verification recipe and payload shape:
llms.txt. Up to 5 active webhooks per account; to
rotate a secret, rm then add again.
Same tools on stdio and remote: notify(text), ask(text, buttons?, timeout?),
check_replies(since?, wait?), check_status(), upgrade().
Local (stdio) — for Claude Desktop / Claude Code:
{
"mcpServers": {
"pingwa": {
"command": "uvx",
"args": ["pingwa", "mcp"],
"env": { "PINGWA_KEY": "pw_your_key" }
}
}
}
Remote (Streamable HTTP, no install):
{
"mcpServers": {
"pingwa": {
"url": "https://pingwa.dev/mcp",
"headers": { "Authorization": "Bearer pw_your_key" }
}
}
}
Every error carries an action hint, so an agent knows what to do next.
Only your message text and your key, to the pingwa backend you point it at — over HTTPS. Nothing else. The recipient is always your own phone (there is no "to" field). See the service's /privacy.
MIT — see LICENSE in the repository root. It covers this client package only
(src/pingwa/client/); the pingwa server is separate, private, and not
published. See NOTICE for the exact boundary.
Source-derived launch command. Check the maintainer’s required arguments and credentials before running:
uvx pingwaMerge this template into ~/Library/Application Support/Claude/claude_desktop_config.json. Keep existing servers. Add any arguments, credentials, and permissions required by the maintainer; this template has not been install-tested.
{
"mcpServers": {
"io-github-grzgrzgrz3-pingwa": {
"command": "uvx",
"args": [
"pingwa"
]
}
}
}Restart Claude Desktop completely for changes to take effect. Confirm the server appears connected in the client’s tool list, then try a read-only example from its documentation.
Claude Desktop setup referencepingwapypipingwa works with any MCP-compatible client. Copy the config snippet from the Configuration section above and add it to the file shown for your client, then restart the application.
~/Library/Application Support/Claude/claude_desktop_config.jsonRestart Claude Desktop completely for changes to take effect.~/.cursor/mcp.jsonRestart Cursor for changes to take effect..vscode/mcp.jsonReload VS Code window for changes to take effect.~/.codeium/windsurf/mcp_config.jsonRestart Windsurf for changes to take effect..mcp.jsonSave at the project root, then start Claude Code in that project and review the MCP server approval prompt. Keep real credentials out of shared files.