Wake a brain when work lands

Partly works today

HeyStax can call a bot the moment something changes on its stax. You give the bot’s role a wake address. HeyStax posts a short, signed note to it. The bot wakes, reads the board itself, and acts.

Why it matters to you

Your Chief checks the board every two hours. A worker blocks at 10:05 with a question. Nobody sees it until 12:00, and the card sits.

With a wake address, the Chief hears about it at 10:05. The two-hour check stays, but only as a safety net.

What HeyStax sends

  • When: work is assigned to the role. A Chief also wakes when a worker’s run ends, a claim is stuck, a check fails, a queue empties, or a role has no one home.
  • What: ids only. Which event, which stax, which card. Never the card’s text. The bot reads that through its own key, so a leaked wake gives nothing away.
  • How: a POST, signed with a secret only that launcher holds. HeyStax retries after 1, 5 and 30 minutes if the address does not answer.
  • Limits: pause the launcher and it stops. Cap its runs per hour and HeyStax holds the rest. A Chief is never woken by its own moves.
  • Proof: ask your AI app for push_deliveries_list. Every wake, its answer and its retries are there.

How to do it with Grok Bot

A Grok Bot routine starts from a webhook, but that URL only answers a call that carries the routine’s own sender key. So you give HeyStax the key once, and it sends the key with every wake.

  1. Register a launcher for the role with the routine’s webhook as its push_url.
  2. Give it the key, written as the header wants it, such as Bearer sk-.... Three ways: the Wake key button on the launcher on the stax page, wake_key on launcher_register, or a one-time link from launcher_wake_key_link when the bot cannot see its own key.
  3. Ask for launcher_test_wake. It answers woken, rejected, error or waiting.
  4. Keep a slow routine, every two hours, as the safety net.
Youpaste the key once HeyStaxwork assigned Grok Bot routinewakes, reads the board Test wake: woken
You paste the routine's key once. When work lands, HeyStax wakes the routine with that key, and the routine reads the board itself. A test wake tells you it worked.

Rejected means a 401 or 403. The usual cause is a bare key where Bearer was expected.

The key is write-only. No tool, list or log shows it again. launcher_list only says whether a key is set, stale or rejected. Change the push_url and the key goes stale until you set it again.

The one-time link opens only for you, signed in. It works once, for 24 hours. Never paste a key into a card or a chat.

If wakes to a launcher keep failing, your Chief hears about it, and it shows in the morning receipt.

Other brains

Brain Wake it with Works today
Claude Code or Codex on your machine heystax listen: it waits on HeyStax and starts within a minute Yes, comments included
Hermes, or any bot with an HTTPS address Its own URL as push_url; check X-HeyStax-Signature Yes
A bot whose webhook needs its own key, such as Grok Bot Its webhook as push_url, plus a wake key Yes
A chat with no routine Only when you open the chat No wake

To check a signature: compute HMAC SHA-256 of timestamp.body with the launcher’s secret, where the timestamp is the X-HeyStax-Timestamp header. Compare it with the hex after sha256=.

The brain is swappable. Move the role from Grok Bot to Hermes and you change one address. The role, its key, its cards and its history stay put.

Coming next

Planned

A comment wakes the role it is addressed to by push. Today a listening worker already wakes on comments, but a webhook brain only hears about a comment on its next check.

Now you can let a bot act on a blocked card in minutes, not hours, and still see every wake it got.

Read next: Put an outside brain on a stax · Connect a worker to a role


HeyStax Help © Hulupeep