Railway Cost Reduction Plan

Updated: February 1, 2026

VERIFIED Analysis (January 13, 2026)

Primary Database: TimescaleDB

The fulcrum-server connects to:

POSTGRES_CONN_STR=postgresql://fulcrum:fulcrum_prod@timescaledb-docker.railway.internal:5432/fulcrum_metrics

TimescaleDB is the ONLY database in use.

Services to DELETE (Confirmed Unused)

Service	Reason	Estimated Savings
Postgres-Wk0k	No applications connect to it	~$10-15/month
Postgres-b6-Y	No applications connect to it	~$10-15/month
clickhouse	Not used in any Go code	~$15-25/month
migrator	One-time service, should be stopped	~$5/month

Total Estimated Savings: $40-60/month

Services to KEEP

Service	Purpose
timescaledb-docker	Primary database (fulcrum_metrics)
Redis	Caching
nats	Message queue
fulcrum-server	Main API + MCP endpoint
event-processor	Event processing
dashboard	Next.js frontend

Note: fulcrum-mcp was deleted 2026-01-31. MCP is now served via fulcrum-server at /mcp.

HOW TO DELETE (Railway Dashboard)

Open: https://railway.app/project/4f064d07-7215-4097-b87c-258184e3febf
For each service to delete:
Click on the service card
Go to Settings tab
Scroll to bottom → Danger Zone
Click Delete Service
Type service name to confirm

Delete in this order:

clickhouse
Postgres-Wk0k
Postgres-b6-Y
migrator (or just stop it if you want to keep for future migrations)

Post-Deletion Verification

After deleting, run these checks:

# Verify fulcrum-server still works
curl https://fulcrum-server-production.up.railway.app/health

# Check dashboard
curl https://fulcrumlayer.io

Additional Optimizations (Optional)

Enable Auto-Sleep

For each app service (fulcrum-server, event-processor, dashboard): - Settings → Deploy → Sleep after inactivity - Set to 5-10 minutes

Scale Down Resources

Pre-production sizing: - fulcrum-server: 256MB RAM - event-processor: 256MB RAM - dashboard: 512MB RAM - Redis: 128MB - NATS: 128MB - TimescaleDB: 512MB

Status: Ready for Execution Verified by: Audit Agent - January 13, 2026