Separate RPC compute from streaming data
Alchemy prices ordinary API work in compute units and Solana gRPC by streamed data. Supanode sells RPC, WebSocket and eligible gRPC access together in Bundles. A single headline monthly price cannot describe these different workloads. Start with the calls and subscriptions your backend actually needs.
Terms checked on 21 October 2026. USD prices preserve each provider's units. Alchemy's free RPC allowance does not make mainnet gRPC free.
Comparable services and their units
| Workload or term | Supanode | Alchemy |
|---|---|---|
| Standard RPC entry | STARTER $40 / 30 days; 15 RPS, 5 TPS | Free: 30M CU / month, 15 requests/s |
| Paid RPC billing | Bundle price within quotas | Pay as You Go: $0.525 / million CU |
| Mainnet Yellowstone gRPC | FOCUS $99 / 30 days; one subscription, 100 watched addresses | $75 / TB streamed; no prerequisite plan or monthly minimum |
| gRPC replay | 100 slots | 6,000 slots through from_slot |
| Starting a test | Up to 24-hour product trial; follow portal instructions | Free RPC tier; gRPC usage is separately priced |
RPC figures: Alchemy pricing. Streaming price and entry conditions: Alchemy Solana gRPC. Replay: Alchemy Yellowstone gRPC overview.
What the Supanode plans include
STARTER is $40 per 30 days for RPC and WebSocket. Mainnet gRPC begins with FOCUS at $99 per 30 days: 25 RPC requests/s, 10 transactions/s, one gRPC subscription and up to 100 watched addresses. BUILD is $159 per 30 days with 200 RPC requests/s, 30 transactions/s, 10 gRPC subscriptions and 5,000 watched addresses in total. Owner/program filter limits also apply.
Supanode gRPC replay covers 100 slots. RPC replay covers two days. These are separate interfaces and windows; neither is a promise of arbitrary historical access. A higher plan buys stated capacity. It does not establish a latency advantage.
For a trial of up to 24 hours, enter the portal, open Free trial and follow the instructions there. The trial duration and setup follow those instructions. See Free trial and the full plan matrix.
Build a workload budget
For Alchemy RPC, count compute units by method rather than treating every request as the same cost. Keep the RPC budget separate from your gRPC data budget. At the published streaming rate, 2 billed TB would be $150 in gRPC usage. This is arithmetic from the rate, not a quote for your combined application or proof that 2 TB fits a Supanode plan.
Supanode can suit a bounded address list when its Bundle quotas match the application and a 100-slot replay window fits the recovery design. Alchemy's metered streaming and documented 6,000-slot replay deserve testing when your feed or outage target is larger. Choose using measured bytes, filters and recovery requirements.
Alchemy describes mainnet and devnet endpoints with X-Token authentication, and its subscription API covers accounts, transactions, slots and blocks. Standard Yellowstone messages can simplify portability. Provider-specific behavior still needs validation in your client. Alchemy Yellowstone gRPC overview and Alchemy subscription API.
Keep live feeds and historical SQL separate
A live Yellowstone feed supplies updates for your own processing pipeline. Supanode's Historical Data Indexer is a separate SQL service with its own price and schema. Alchemy data APIs are also distinct from a raw gRPC subscription. Compare the output and history you require before substituting one product for another.
If your application already depends on an Alchemy-specific API, inventory that dependency before changing the endpoint. The presence of a standard RPC service does not establish a replacement for every managed API or every chain.
Migration checklist
- List standard Solana calls separately from Alchemy-specific APIs, webhooks and chain dependencies. Identify which calls can move unchanged and which need an adapter.
- Export RPC compute-unit usage and gRPC data usage separately. Use the same measurement period and retain the original billing units.
- Configure the destination endpoint, token and pinned Yellowstone client. Compare filters, account data slices, commitment levels and decoded outputs.
- Test a restart from a saved slot, including an outage longer than 100 slots. Confirm your separate backfill path covers any gap.
- Run a small production read cohort, compare completeness and errors, then shift traffic with a rollback endpoint available.