Weekly competitor pricing tracker logged to Google Sheets

Build a code workflow that tracks competitor pricing on a weekly cadence and logs every change to a Google Sheet.

Trigger: cron, every Monday at 9am in the user's local timezone.

Setup inputs: (1) a list of competitor pricing page URLs, each with a short competitor name label, (2) a Google Sheets spreadsheet ID and the tab name to write to.

Step 1 — Scrape each competitor pricing page. Iterate over the configured URLs and call Firecrawl's Scrape URL operation for each one. Use the JSON extraction format with a structured schema that pulls an array of plans, where each plan has: plan_name (string), price_amount (number), price_currency (string, default USD), billing_period (string, one of monthly, annual, one-time, contact-sales). Pass a short natural-language prompt alongside the schema so Firecrawl knows to extract the standard pricing tiers and skip add-ons and enterprise contact rows. Collect the results into a single array shaped as { competitor, url, scraped_at, plans: [...] }.

Step 2 — Load the previous week's snapshot from Google Sheets. Call the Get Values operation on the configured spreadsheet and tab, reading the full data range. The sheet's columns are: run_date, competitor, plan_name, price_amount, price_currency, billing_period, change. Use the most recent run_date in the sheet as the prior snapshot to diff against.

Step 3 — Run the diff in a Daytona sandbox. Call Create Sandbox to provision a fresh Python sandbox. Then call Run Code in Sandbox with a Python script that takes the current scrape array and the prior snapshot as inputs and produces, for each (competitor, plan_name) pair, a normalized row with a change column set to one of: new (plan didn't exist last week), price_up (same plan, higher price), price_down (same plan, lower price), unchanged (same plan, same price), removed (plan existed last week but is missing this week). Match plans by (competitor, plan_name) case-insensitively. Output the result as a JSON array of rows. After the script returns, call Delete Sandbox to clean up — every run should leave zero sandboxes behind.

Step 4 — Append the new snapshot to Google Sheets. Call the Append Values operation to add one row per plan from the diff output. Each row is: today's date, competitor, plan_name, price_amount, price_currency, billing_period, change. Use USER_ENTERED so dates and numbers render correctly.

Error handling: if Firecrawl fails to extract from a single URL, log an error row for that competitor with change set to 'scrape_failed' but still process the others. If the Daytona sandbox creation or code execution fails, attempt Delete Sandbox in a cleanup step regardless so we never leak sandboxes.

This is a deterministic pipeline. Every step has fixed inputs and outputs, no judgement calls or LLM reasoning required inside the workflow itself.