Help

Quick start

1. Sign in and create a job (or use Create with Chat from the Dashboard).
2. Write a prompt template and optionally add Variables (JSON).
3. Run Preview and (optionally) enable test-send to validate delivery.
4. Pick a schedule (daily/weekly/cron), choose a delivery target (in-app is simplest), then Save Job.
5. Monitor Run History for status, errors, and Sources when web search is enabled.

Templates and variables

• Use placeholders like {{company}} inside your prompt template.
• Variables must be a JSON object with string values (e.g. {"company":"Acme"}).
• Built-ins: {{now_iso}}, {{date}}, {{time}}, {{timezone}}.
• Unknown placeholders are left as-is, which helps catch typos.

Post prompt (optional)

• Enable it to run a second prompt after the model responds, before delivery.
• Use {{output}} to reference the primary output.
• If enabled but blank, it's treated as disabled (you'll see a warning).

Channel setup

• In-app: no setup. Runs are stored in Run History.
• Discord: provide a webhook URL.
• Telegram: provide a bot token and chat ID.
• Custom webhook: provide a URL and method. Headers JSON is optional. Payload JSON is optional.

Custom webhook

If Payload is empty, Promptloop sends a stable default JSON object. If you provide Payload, it will be sent as-is (must be valid JSON).

{"Authorization":"Bearer <token>"}

{
  "title": "[Job Name] 2026-02-20 09:00",
  "body": "<model output>",
  "content": "<full message>",
  "usedWebSearch": false,
  "citations": [{ "url": "https://example.com", "title": "Example" }],
  "meta": { "jobId": "<id>", "runHistoryId": "<id>" }
}

• If your endpoint expects JSON, include a Content-Type header (often `application/json`).
• For GET requests, no request body is sent (payload is ignored).
• Use Preview with test-send enabled to validate delivery before saving.

Preview and test-send

Preview runs the prompt immediately. Enable test-send to send preview output to the selected channel before saving changes.

Web search (optional)

• Turn on web search only when you need fresh facts or live sources.
• When web search is enabled, runs include Sources links (and may take longer).
• If web search is enabled but no sources are found, the run fails so you can adjust the prompt.

Run history

• Run History shows the latest executions with status and a short preview of output.
• Failures include a short error message to help you fix prompts, schedules, or channel settings.
• Web-search runs show Sources so you can audit where facts came from.

Need help?

• Start with Run History: it shows the latest error messages and (when enabled) Sources.
• For webhook issues, inspect your endpoint logs and validate the request body/headers.
• If you are self-hosting, check server logs for /api/cron/run-jobs and delivery errors.