Get a text when your build finishes

A completion text should tell you what happened

A build that takes forty minutes should not require forty minutes of watching a terminal. With a connected agent, you can ask for a short SMS when the result is ready and decide whether you need to return to the computer.

Include the task name, whether it succeeded, and anything that needs attention. “api-tests passed; two warnings changed” gives you more to work with than “done.” Save full logs and detailed artifacts in the workspace where they can be reviewed.

The same pattern works for an agent supervising a test suite, evaluation, data import, or render. The agent needs to observe the actual result before sending a completion message.

Connect and test before the long run

1. Install the local connector in the environment running your agent. Use Codex or Claude Code instructions for those clients.
2. Ask the agent to open a bridge and send a harmless test message.
3. Reply from your verified phone and confirm the agent retrieves and answers it.
4. Give the agent the build task and your completion-message rule.

Keep the agent running and its computer awake and online. Campfire SMS does not monitor a standalone process just because the connector is installed. A CI job needs an integration or agent that observes the job result and sends the text.

Give the build a clear notification rule

Run the agreed build and checks. Use Campfire SMS to text me once you have the final result, or earlier if progress needs my input. Include the task name, pass or fail, and a short summary of changed warnings. Save full output in the workspace. Check my replies at meaningful work boundaries and answer within our existing task scope. Close the bridge when we are finished.

For a job that can take hours, tell the agent whether it should continue useful work while waiting or remain available for a bounded reply window. Choose that behavior explicitly so a completion text does not leave you assuming the session will stay open indefinitely.

If you want a timer-based follow-up the next morning after this session ends, use a persistent reminder workflow.

Reply with the next useful question

• “Which checks failed?” asks for a concise failure summary.
• “Only summarize warnings introduced by this change.” keeps the response focused.
• “Show me the slowest three checks.” asks the agent to analyze output it already has.
• “Prepare a PR summary for review.” moves from the result toward a review-ready update.

These replies carry task input back to the connected agent. Actions still use its existing tools and permissions; a native approval can require the host's own interface.

If the build finishes and no text arrives

Confirm that the agent observed completion, kept its bridge open, and actually called the send tool. Then check the computer's sleep state, internet connection, installation status, and account message balance.

If the text arrived but your follow-up gets no answer, inspect whether the agent is still active and checking its inbox. Outbound delivery and processing a reply are separate steps. The MCP messaging guide explains the full exchange.

Common questions