A practical field guide from Automation Ace.
The short answer
For running automation code outside Zapier, pick the serverless platform that matches your job's duration and where your data lives. Cloudflare Workers excel at fast webhook endpoints with minimal cold starts and a generous free tier. AWS Lambda handles heavier jobs up to 15 minutes and AWS-centric stacks. Vercel Functions fit teams already deploying apps on Vercel. Supabase Edge Functions fit apps built on Supabase's database and auth.
Part of the automation maturity ladder (stage 3). Why move code out of Zapier at all? See 5 reasons to move custom code to serverless.
The use case: automation code outside Zapier
Typical jobs: receive and verify webhooks, transform data, call APIs with retries, run scheduled syncs, and return results to a Zap or app. The criteria that matter most are cold starts (for webhook response time), maximum run time, free-tier fit, scheduling, and proximity to your data.
Comparison at a glance
| AWS Lambda | Cloudflare Workers | Vercel Functions | Supabase Edge Functions | |
|---|---|---|---|---|
| Runtime | Node.js, Python, Java, .NET, Ruby, custom | V8 isolates: JavaScript, TypeScript, Python, Wasm | Node.js, Python, and more (fluid compute) | Deno: TypeScript and JavaScript |
| Cold starts | Noticeable for some runtimes; mitigations available | Minimal (isolates start very fast) | Reduced with fluid compute; still possible | Generally small; some on first invocation |
| Max run time | Up to 15 minutes | Short CPU limits per request; waiting on I/O does not count toward CPU | Hobby up to 300s with fluid compute (60s without); longer on paid plans | 150s wall clock on free, 400s on paid; about 2s CPU |
| Free tier (check current pricing) | Historically 1M requests and 400,000 GB-seconds per month | 100,000 requests per day | Hobby plan; personal, non-commercial use | Included invocations on the free plan |
| Scheduling | EventBridge Scheduler | Cron Triggers | Cron jobs | pg_cron or external schedulers |
| Built-in data | AWS services (S3, DynamoDB, SQS) | KV, D1, R2, Queues, Durable Objects | Pairs with Vercel storage and Marketplace integrations | Supabase Postgres, Auth, Storage |
| Best for | Heavier and longer jobs; AWS-centric stacks | Fast webhooks and edge endpoints; low cost at volume | Teams already deploying web apps on Vercel | Apps built on Supabase |
Limits, free tiers, and plan terms change frequently; values reflect each vendor's documentation at the time of writing. Always confirm before committing.
Cold starts
A cold start is the extra delay when a platform spins up a new instance. For webhooks that must answer quickly, isolate-based platforms like Cloudflare Workers have an edge. Lambda cold starts vary by runtime and package size, with mitigations such as provisioned concurrency at extra cost. For scheduled and background jobs, cold starts rarely matter. See why webhooks fire twice for why slow responses cause retries.
Timeouts and long jobs
Match the job to the limit: quick glue on Workers or Edge Functions, multi-minute processing on Lambda or Vercel's longer durations. For anything that waits hours or days or needs per-step retries, add a durable engine; see durable workflows explained and long-running automations.
Pricing and free tiers
- Cloudflare Workers: a daily free request allowance and a low-cost paid plan; see free-plan automation ideas.
- AWS Lambda: a long-standing monthly free tier for requests and compute, but related services (API Gateway, logs, data transfer) bill separately.
- Vercel: the Hobby plan is intended for personal, non-commercial use; business automations generally belong on a paid plan.
- Supabase: Edge Functions invocations are included on the free plan with limits, alongside the database.
Compare with your Zapier task spend using the automation ROI calculator and tasks vs operations pricing.
How to choose
| If your automation... | Start with |
|---|---|
| Webhook receivers and lightweight glue | Cloudflare Workers |
| Jobs that run several minutes, or heavy processing | AWS Lambda |
| Data already in AWS | AWS Lambda |
| Your web app is on Vercel and the automation belongs with it | Vercel Functions (on a plan that permits commercial use) |
| Your data and auth live in Supabase | Supabase Edge Functions |
| Long multi-step processes with waits | Any of these plus a durable engine |
Design for portability
- Keep business logic in plain functions, separate from platform-specific handlers.
- Use standard
fetchand JSON wherever possible. - Store secrets in each platform's secret manager, never in code.
- Make handlers idempotent and log request IDs.
- Keep infrastructure in version control. See version control for automations.
Whichever platform you choose, the trigger pattern matters as much as the host; see event-driven vs scheduled vs request-response.
Frequently asked questions
Which serverless platform is best for automation code?
It depends on the job: Cloudflare Workers for fast webhook endpoints and low cost at volume, AWS Lambda for heavier jobs up to 15 minutes and AWS stacks, Vercel Functions for teams already deploying apps on Vercel, and Supabase Edge Functions for apps built on Supabase.
Which serverless platform has the fastest cold starts?
Isolate-based platforms such as Cloudflare Workers generally have minimal cold starts. Lambda cold starts vary by runtime and package size, and Vercel's fluid compute reduces cold starts.
How long can serverless functions run?
At the time of writing, AWS Lambda allows up to 15 minutes, Vercel Hobby functions up to 300 seconds with fluid compute, Supabase Edge Functions 150 seconds of wall-clock time on free plans and 400 on paid, and Cloudflare Workers enforce CPU-time limits rather than long wall-clock runs. Check current documentation.
Can I use Vercel's free plan for business automations?
Vercel's Hobby plan is intended for personal, non-commercial use, so business automations generally belong on a paid plan.
Should I use serverless functions or a durable workflow engine?
Use serverless functions for short tasks like webhooks, transforms, and API calls. Add a durable engine when processes wait hours or days, span many steps, or need per-step retries.
Disclaimer: Templates and checklists are starting points; adapt them to your process, policies, and tools. Product features, limits, and pricing mentioned here change over time, so confirm current details in each vendor's documentation. This article may include links to apps, products, or services; some links may be affiliate links, which means Automation Ace may earn a commission at no extra cost to you.