Problem
I’m encountering a deployment error that appears to be a platform-side bug rather than a project configuration issue.
Error: No more than 12 Serverless Functions can be added to a Deployment on the Hobby plan. Create a team (Pro plan) to deploy more.
Project Setup
- Framework setting:
null(not a framework-based project) - Runtime:
@vercel/node - Middleware: No middleware file
- Config: Single
vercel.jsonconfiguration
Investigation & Findings
I ran npx vercel@latest build locally, which completed successfully. The generated .vercel/output/functions/api directory contains exactly 11 function bundles:
/api/finny
/api/goals
/api/memory
/api/plaid
/api/plaid_management
/api/process-notifications
/api/refresh_financial_data
/api/scheduled-sync
/api/store_accounts
/api/transactions_sync
/api/webhook
The .vercel/output/builds.json confirms this: 11 @vercel/node builds + 1 @vercel/static build.
Key Evidence of Platform-Side Issue
I tested deploying with npx vercel@latest deploy --prebuilt --force, which deploys the pre-generated .vercel/output directory directly. The deployment still fails with the same function-count error.
Since --prebuilt bypasses any build-time function generation and uses exactly what’s in the output folder, this suggests the error is occurring during Vercel’s deploy step — not because my repo is generating hidden functions.
Additionally, I verified via Vercel support chat that the deployment metadata shows lambdaRuntimeStats: {"nodejs": 11}, meaning Vercel is correctly detecting 11 functions, yet the deployment is still being rejected as if it exceeds 12.
What I’ve Ruled Out
No hidden files in the apifolder
No SSR/ISR pages (framework is null)
No middleware
No dynamic routes generating additional functions
Local build output matches expected 11 functions
Expected vs Actual Behavior
- Expected: Deployment should succeed with 11 functions (under the 12 function Hobby limit)
- Actual: Deployment fails claiming I’ve exceeded the limit
Request
Could someone from the Vercel team investigate what’s happening during the deploy step? The evidence suggests a platform-side miscount or regression in the function limit validation logic.
Need this fixed ASAP