Create a Parallel Monitor by describing what you want to track in plain English. Hermes helps refine the request, previews the scope, shows the expected cost, and sends new Monitor events to Telegram, Slack, Discord, or another connected destination.
Beta: This integration is ready for testing, but you should send a controlled test notification before relying on it for important alerts.
You need:
- Hermes Agent with a persistent messaging gateway
parallel-cliversion 0.7.1 or later- Stored Parallel credentials created with
parallel-cli login - A connected destination such as Telegram, Slack, or Discord
Run:
parallel-cli --version
parallel-cli login
hermes gateway statusFrom your terminal:
hermes skills install georgeatparallel/hermes-parallel-monitor/skills/parallel-monitor --yesThen start a new Hermes conversation so it loads the skill.
You can also install it inside a Hermes conversation:
/skills install georgeatparallel/hermes-parallel-monitor/skills/parallel-monitor --now
For example:
Keep me updated when OpenAI, Anthropic, or Google announces a new model or major API feature. Send me a daily Telegram digest at 9am.
Hermes will:
- Ask any important clarification questions.
- Show a Starting snapshot so you can check the scope.
- Show the exact query, schedule, destination, and estimated 30-day cost.
- Wait for you to explicitly approve creation.
- Create and smoke-test the Monitor and its notification delivery.
Nothing billable is created until you approve the final configuration.
After creation, Hermes will offer to send a labeled test notification. Approve it if you want to verify that the message reaches the correct destination.
The test proves notification routing. Real Monitor notifications arrive later when Parallel detects a relevant change.
Alert me in Telegram when Acme changes its pricing page.
Send me a daily digest of new AI developer-tool launches at 8am Pacific time.
Watch for new SEC filings from Tesla and notify me when one appears.
You can manage everything conversationally:
What am I monitoring?
Pause notifications for my Acme pricing Monitor.
Change my AI launches digest to 9am.
Permanently stop my Acme pricing Monitor.
Hermes always asks for confirmation before a billable off-schedule run or permanent cancellation.
- Creating a Monitor starts one execution immediately and future executions are billable at the confirmed cadence.
- Pausing Hermes notifications does not pause the Parallel Monitor or its cost.
- Quiet alerts send nothing when there is no new event.
- Daily and weekly digests can optionally send a no-change message.
- The Starting snapshot is only a preview. Those results are not replayed as future notifications.
- Delivery is at-least-once, so an unusual failure may produce a duplicate notification.
- The Hermes gateway must remain running for scheduled delivery.
Update the skill with:
hermes skills update parallel-monitorIf Hermes reports that the copied notification runtime is outdated, open a new conversation and ask:
Upgrade my Parallel Monitor notification runtime.
Hermes will pause the affected notification jobs, migrate and smoke-test the runtime, and resume them only after verification succeeds.
Ask Hermes to permanently stop the Monitor and confirm when prompted. Hermes will cancel the server-side Monitor, remove its notification job, and delete its local notification state.
For manual verification, set RUNTIME to the installed skill's scripts/hermes_monitor_runtime.py path and run the unscoped status command:
python3 "$RUNTIME" statusVerify that the removed Monitor ID is absent from the unscoped status response's monitors list. Uninstall the skill only after every Monitor you no longer want has been removed.
First ask Hermes to permanently stop any Monitors you no longer want. Then uninstall the skill:
hermes skills uninstall parallel-monitorUninstalling the skill alone does not cancel server-side Parallel Monitors or stop their usage.
- The skill requires credentials stored by
parallel-cli login. - It does not copy API keys into cron jobs, scripts, or Monitor state.
- Monitor content is treated as untrusted web data and sanitized before delivery.
- Billable and destructive actions require explicit confirmation.
Run the test suite from the repository root:
python3 -m unittest discover -s testsThe installed runtime uses only the Python standard library. The test suite covers the released CLI contract, creation and recovery, scheduling, delivery acknowledgement, cleanup, and skill packaging.