To choose a Solana RPC provider, start from your workload: reads, heavy account queries, streaming or sending transactions. Check documented rate limits, how transactions land, regions, support and uptime history, then benchmark two providers with your own traffic for a week. Use a second provider as a fallback, and never expose an unrestricted key in the frontend.
| Shared plan | Dedicated node | |
|---|---|---|
| Cost | Lower, scales with requests | Much higher, fixed |
| Performance | Varies with other customers | Steady, reserved for you |
| Heavy queries | Often limited or priced up | Usually fewer limits |
| Good for | Early stage and most apps | Trading, indexers, high load |
Plans and prices differ by provider: ask for current limits in writing.
- The public Solana endpoints are rate limited and are not for production.
- Match the provider to your workload: plain reads, heavy account queries, streaming or transaction sending behave very differently.
- Ask how transactions land and what happens at the rate limit, not just for a requests-per-second number.
- Benchmark with your own traffic and compare latency, errors and slot lag.
- Use two providers, keep API keys out of the frontend, and test the failover.
An RPC provider sits between your app and the chain. When it is slow or drops requests, users blame your product, not the node. Choosing one is mostly a question of what you ask it to do, and a few checks before you sign save a painful migration later.
Start with your workload
- Reads for a frontend: balances, accounts, recent transactions. Needs low latency near your users and steady availability.
- Heavy account queries: calls such as getProgramAccounts return a lot of data and are often limited or priced differently.
- Streaming and indexing: WebSocket subscriptions or a gRPC stream (Geyser-based) for live data. Check what is included and what is extra.
- Sending transactions: bots, keepers and launches live or die on how reliably transactions reach the leader.
Ten questions to ask
- 1. What are the exact rate limits? per second, per method and per plan, and what happens when you hit them: errors, throttling or a surprise bill.
- 2. How do you land transactions? ask about stake-weighted connections, retries and priority fee support, and request landing statistics for your use case.
- 3. Which regions do you serve? pick one close to your backend and your users, and check if you can run in two regions.
- 4. What streaming options exist? WebSockets, gRPC and what they cost.
- 5. How much history do you keep? full history, archival access and indexing add cost: say what you need.
- 6. What is the uptime record? a public status page with past incidents is a better signal than a number in a sales deck.
- 7. What support do you give? a named channel, response times, and whether anyone answers at the hour of your launch.
- 8. How are keys secured? IP or domain allowlists, rotation and per-key limits, so a leaked key cannot be abused.
- 9. How do you price overages? ask for an example bill for a traffic spike.
- 10. Can we leave easily? standard endpoints and no lock-in to a proprietary API for core calls.
Test before you commit
A benchmark of simple calls tells you little. Run your own traffic against two providers for a week, including your heaviest calls and some peak hours. Record median and slow latency, error rates, rate-limit responses and how far each reports its slot behind a second source. Ask for a trial or a short plan to do it.
Security and red flags
- Keys in the frontend: an unrestricted key in browser code will be copied. Restrict by domain or proxy requests through your backend.
- "Unlimited" with no terms: ask what unlimited means in the fair-use policy.
- No incident history or named support: you cannot judge reliability you cannot see.
- Pushing a dedicated node early: most early-stage apps do well on a shared plan with a fallback.
For terms and fees around priority transactions see the glossary entries for RPC and priority fee. To compare providers that Orivions has verified, see Solana infrastructure providers.
Also useful: how to monitor a Solana program in production.
Frequently asked questions
Can I use the public Solana RPC endpoint in production?
No. The public endpoints are rate limited and are meant for development and light use. A production app needs a provider with documented limits, support and an agreement on uptime.
What is the difference between shared and dedicated Solana RPC nodes?
Shared plans split capacity between customers and are cheaper, but a busy neighbour or a traffic spike can slow you down. A dedicated node is reserved for you, costs much more, and gives steady performance and fewer limits on heavy calls.
Do I need more than one RPC provider?
For anything that holds user funds or runs keepers, yes. One provider is a single point of failure. Configure a second one as a fallback and test the switch.
How do I test a Solana RPC provider before committing?
Run your own real workload against it for a week and record median and slow latency, error rate, rate-limit responses and how far its reported slot lags behind a second source. A benchmark of simple calls tells you little about your use case.
Find a verified Solana infrastructure provider
Describe your project and load. Smart Match suggests verified infrastructure providers with a reason for each match. Free for teams.
Get matched →Looking for help with this? Browse Solana infrastructure and RPC providers, or describe your project and get matched for free.
Related reading
- Solana program monitoring: what to watch and alert onWhat to monitor on a live Solana program: failed transactions, upgrade and authority changes, keeper balances, oracle freshness and RPC lag, plus how to get the data and who should own each alert.
- Solana indexer vs RPC: what to use, and when to add an indexerWhen RPC is enough on Solana and when you need an indexer: history, filters and totals, Geyser streaming, decoding, build or buy, and questions to ask a provider.
- Solana program upgrade authority: how to secure it with a multisigWho should hold a Solana program upgrade authority, how to move it to a multisig, when to add a timelock or make the program immutable, and the mistakes to avoid.